`r` is moved-from, and later captured in a different lambda. The compiler may
choose to move and perform the other capture later, resulting in a use-after-free.
Fix by copying `r` instead of moving it.
Discovered by sstable_test in debug mode.
Message-Id: <20170702082546.20570-1-avi@scylladb.com>
(cherry picked from commit 07b8adce0e)
* seastar e23b9b8...a66e0c5 (3):
> posix.hh: add missing include
> tls_test: Fix echo test not setting server trust store
> tls: Actually verify client certificate if requested
Fixes#3072
"_free_segments_in_zones is not adjusted by
segment_pool::reclaim_segments() for empty zones on reclaim under some
conditions. For instance when some zone becomes empty due to regular
free() and then reclaiming is called from the std allocator, and it is
satisfied from a zone after the one which is empty. This would result
in free memory in such zone to appear as being leaked due to corrupted
free segment count, which may cause a later reclaim to fail. This
could result in bad_allocs.
The fix is to always collect such zones.
Fixes#3129
Refs #3119
Refs #3120"
* 'tgrabiec/fix-free_segments_in_zones-leak' of github.com:scylladb/seastar-dev:
tests: lsa: Test _free_segments_in_zones is kept correct on reclaim
lsa: Expose max_zone_segments for tests
lsa: Expose tracker::non_lsa_used_space()
lsa: Fix memory leak on zone reclaim
(cherry picked from commit 4ad212dc01)
Currently scylla-housekeeping-daily.service/-restart.service hardcoded
"--repo-files '/etc/yum.repos.d/scylla*.repo'" to specify CentOS .repo file,
but we use same .service for Ubuntu/Debian.
It doesn't work correctly, we need to specify .list file for Debian variants.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1513385159-15736-1-git-send-email-syuu@scylladb.com>
(cherry picked from commit c2e87f4677)
We have had an issue recently where failed SSTable writes left the
generated SSTables dangling in a potentially invalid state. If the write
had, for instance, started and generated tmp TOCs but not finished,
those files would be left for dead.
We had fixed this in commit b7e1575ad4,
but streaming memtables still have the same isse.
Note that we can't fix this in the common function
write_memtable_to_sstable because different flushers have different
retry policies.
Fixes#3062
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <20171213011741.8156-1-glauber@scylladb.com>
(cherry picked from commit 1aabbc75ab)
Broken after 0ac2c388b6, which assigns
empty reader to _delegate on hitting wide partition limit. The test
assumed that the original _delegate will be invoked when the
single-partition reader is asked for the next partition, which is no
longer the case.
Message-Id: <20170912172739.6851-1-tgrabiec@scylladb.com>
The infrastructure this test relies on is not present in 1.7, so
just remove the test as backporting the required changes would be a
risky, non-trivial effort.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20170912081304.10116-1-duarte@scylladb.com>
"A new feature flag CORRECT_COUNTER_ORDER is introduced to allow seamless
upgrade from 1.7.4 to later Scylla versions. If that feature is not
available Scylla still writes sstables and sends on-wire counters using
the old ordering so that it can be correctly understood by 1.7.4, once
the flag becomes available Scylla switches to the correct order.
Fixes #2752."
* tag 'fix-upgrade-with-counters-1.7/v1' of https://github.com/pdziepak/scylla:
tests/counter: verify counter_id ordering
counter: check that utils::UUID uses int64_t
mutation_partition_serializer: use old counter ordering if necessary
mutation_partition_view: do not expect counter shards to be sorted
sstables: write counter shards in the order expected by the cluster
tests/sstables: add storage_service_for_tests to counter write test
tests/sstables: add test for reading wrong-order counter cells
sstables: do not expect counter shards to be sorted
storage_service: introduce CORRECT_COUNTER_ORDER feature
tests/counter: test 1.7.4 compatible shard ordering
counters: add helper for retrieving shards in 1.7.4 order
tests/counter: add tests for 1.7.4 counter shard order
counters: add counter id comparator compatible with Scylla 1.7.4
tests/counter: verify order of counter shards
tests/counter: add test for sorting and deduplicating shards
counters: add function for sorting and deduplicating counter cells
counters: add more comparison operators
Until the cluster is fully upgraded from a version that uses the
incorrect counter shard ordering it is essential to keep using it lest
the old nodes corrupt the data upon receiving mutations with a counter
shard ordering they do not expect.
If the feature signaling that we have switched to the correct ordering
of counter shards is not enabled it means that the user still can do a
rollback to a version that expects wrong ordering. In order to avoid any
disasters when that happens write sstables using the 1.7.4 order until
we know for sure that it is no longer needed.
Scylla 1.7.4 used incorrect ordering of counter shards. In order to fix
this problem a new feature is introduced that will be used to determine
when nodes with that bug fixed can start sending counter shard in the
correct order.
Due to a bug in an implementation of UUID less compare some Scylla
versions sort counter shards in an incorrect order. Moreover, when
dealing with imported correct data the inconsistencies in ordering
caused some counter shards to become duplicated.
Initialize the system_auth and system_traces keyspaces and their tables after
the Node joins the token ring because as a part of system_auth initialization
there are going to be issues SELECT and possible INSERT CQL statements.
This patch effectively reverts the d3b8b67 patch and brings the initialization order
to how it was before that patch.
Fixes#2273
Signed-off-by: Vlad Zolotarov <vladz@scylladb.com>
Message-Id: <1500417217-16677-1-git-send-email-vladz@scylladb.com>
(cherry picked from commit e98adb13d5)
Fixes#2723.
* tag 'asias/repair_issue_2723_v1' of github.com:cloudius-systems/seastar-dev:
repair: Do not allow repair until node is in NORMAL status
gossip: Add is_normal helper
(cherry picked from commit 2f41ed8493)
segment_zone::migrate_all_segments() was trying to migrate all segments
inside a zone to the other one hoping that the original one could be
completely freed. This was an attempt to optimise for throughput.
However, this may unnecesairly hurt latency if the zone is large, but
only few segments are required to satisfy reclaimer's demands.
Message-Id: <20170410171912.26821-1-pdziepak@scylladb.com>
(cherry picked from commit 0318dccafd)
global_schema_ptr ensures that schema object is replicated to other
cores on access. It was replicating the "synced" state as well, but
only when the shard didn't know about the schema. It could happen that
the other shard has the entry, but it's not yet synced, in which case
we would fail to replicate the "synced" state. This will result in
exception from mutate(), which rejects attempts to mutate using an
unsynced schema.
The fix is to always replicate the "synced" state. If the entry is
syncing, we will preemptively mark it as synced earlier. The syncing
code is already prepared for this.
Refs #2617.
Message-Id: <1500555224-15825-1-git-send-email-tgrabiec@scylladb.com>
(cherry picked from commit 65c64614aa)
index's file output stream uses write behind but it's not closed
when sstable write fails and that may lead to crash.
It happened before for data file (which is obviously easier to
reproduce for it) and was fixed by 0977f4fdf8.
Fixes#2673.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170807171146.10243-1-raphaelsc@scylladb.com>
(cherry picked from commit dddbd34b52)
"This series ensures the always write correct cell names to promoted
index cell blocks, taking into account the eoc of range tombstones.
Fixes#2333"
* 'pi-cell-name/v1' of github.com:duarten/scylla:
tests/sstable_mutation_test: Test promoted index blocks are monotonic
sstables: Consider eoc when flushing pi block
sstables: Extract out converting bound_kind to eoc
(cherry picked from commit db7329b1cb)
It was meant to be run in the foreground since it is waited upon during
stop(), but as it is now from the stop() perspective it is completed
after first connection is accepted.
Fixes#2652
Message-Id: <20170801125558.GS20001@scylladb.com>
(cherry picked from commit 1da4d5c5ee)
database::make_sstable_reader() creates a reader which will need to
obtain a semaphore permit when invoked, so that there is a limit on
sstable read concurrency (edeef03). Therefore, each read may create at
most one such reader in order to be guaranteed to make
progress. Otherwise, the creation of the second reader may deadlock
(in case of system tables) or timeout (non-system tables), if enough
number of such readers tries to do the same thing at the same time.
One instance of the problem fixed by this patch is in cache populating
reader (98c12dc) when we reach partition size limit
(max_cached_partition_size_in_kb). In that case population is
abandoned and a second read is created, while still keeping the old
one alive. We saw this causing deadlocks during schema tables parsing
when system.schema_columns contained large partitions. Fixes#2623.
Another case when this can potentially happen is when populating
readers are recreated by cache. We replace the reader there, but using
assignment, so the old reader is still alive when the new one is
created. This patch fixes two out of three of such cases. The third
one (in a scanning read) is not that easy to fix. That problem doesn't
exist in version 2.0 and master, where the cache is reworked for row
granularity.
Refs #2644.
Message-Id: <1501160300-18097-1-git-send-email-tgrabiec@scylladb.com>
When we install scylla metapackage with version (ex: scylla-1.7.1),
it just always install newest scylla-server/-jmx/-tools on the repo,
instead of installing specified version of packages.
To install same version packages with the metapackage, limited dependencies to
current package version.
Fixes#2642
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <20170726193321.7399-1-syuu@scylladb.com>
(cherry picked from commit 91a75f141b)
We need to consider the _live_endpoints size. The nr_live_nodes should
not be larger than _live_endpoints size, otherwise the loop to collect
the live node can run forever.
It is a regression introduced in commit 437899909d
(gossip: Talk to more live nodes in each gossip round).
Fixes#2637
Message-Id: <863ec3890647038ae1dfcffc73dde0163e29db20.1501026478.git.asias@scylladb.com>
(cherry picked from commit 515a744303)
Allow reading exactly desired byte ranges and fast_forward_to
In commit c63e88d556, support was added for
fast_forward_to() in data_consume_rows(). Because an input stream's end
cannot be changed after creation, that patch ignores the specified end
byte, and uses the end of file as the end position of the stream.
As result of this, even when we want to read a specific byte range (e.g.,
in the repair code to checksum the partitions in a given range), the code
reads an entire 128K buffer around the end byte, or significantly more, with
read-ahead enabled. This causes repair to do more than 10 times the amount
of I/O it really has to do in the checksumming phase (which in the current
implementation, reads small ranges of partitions at a time).
This patch has two levels:
1. In the lower level, sstable::data_consume_rows(), which reads all
partitions in a given disk byte range, now gets another byte position,
"last_end". That can be the range's end, the end of the file, or anything
in between the two. It opens the disk stream until last_end, which means
1. we will never read-ahead beyond last_end, and 2. fast_fordward_to() is
not allowed beyond last_end.
2. In the upper level, we add to the various layers of sstable readers,
mutation readers, etc., a boolean flag mutation_reader::forwarding, which
says whether fast_forward_to() is allowed on the stream of mutations to
move the stream to a different partition range.
Note that this flag is separate from the existing boolean flag
streamed_mutation::fowarding - that one talks about skipping inside a
single partition, while the flag we are adding is about switching the
partition range being read. Most of the functions that previously
accepted streamed_mutation::forwarding now accept *also* the option
mutation_reader::forwarding. The exception are functions which are known
to read only a single partition, and not support fast_forward_to() a
different partition range.
We note that if mutation_reader::forwarding::no is requested, and
fast_forward_to() is forbidden, there is no point in reading anything
beyond the range's end, so data_consume_rows() is called with last_end as
the range's end. But if forwarding::yes is requested, we use the end of the
file as last_end, exactly like the code before this patch did.
Importantly, we note that the repair's partition reading code,
column_family::make_streaming_reader, uses mutation_reader::forwarding::no,
while the other existing reading code will use the default forwarding::yes.
In the future, we can further optimize the amount of bytes read from disk
by replacing forwarding::yes by an actual last partition that may ever be
read, and use its byte position as the last_end passed to data_consume_rows.
But we don't do this yet, and it's not a regression from the existing code,
which also opened the file input stream until the end of the file, and not
until the end of the range query. Moreover, such an improvement will not
improve of anything if the overall range is always very large, in which
case not over-reading at its end will not improve perforance.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20170718110643.8667-1-nyh@scylladb.com>
It is useful for larger cluster with larger gossip message latency. By
default the fd_max_interval_ms is 2 seconds which means the
failure_detector will ignore any gossip message update interval larger
than 2 seconds. However, in larger cluster, the gossip message udpate
interval can be larger than 2 seconds.
Fixes#2603.
Message-Id: <49b387955fbf439e49f22e109723d3a19d11a1b9.1500278434.git.asias@scylladb.com>
(cherry picked from commit adc5f0bd21)
"This series backports all the repair related fixes to enterprise branch and
updates the scylla_repair to send ranges to repair to all the shards in
parallel, indepedently.
With this series, repair can utilize all the CPUs and is much more efficent."
* tag 'asias/repair-backport-branch-1.7.3-v1' of github.com:cloudius-systems/seastar-dev:
repair: Use selective_token_range_sharder
tests: Add test_selective_token_range_sharder
dht: Add selective_token_range_sharder
repair: further limit parallelism of checksum calculation
repair: Do not store the failed ranges
repair: Prefer nodes in local dc when streaming
repair: Repair on all shards
repair: Allow one stream plan in flight
In storage_proxy we arrange the mutations sent by the replicas in a
vector of vectors, such that each row corresponds to a partition key
and each column contains the mutation, possibly empty, as sent by a
particular replica.
There is reconciliation-related code that assumes that all the
mutations sent by a particular replica can be found in a single
column, but that isn't guaranteed by the way we initially arrange the
mutations.
This patch fixes this and enforces the expected order.
Fixes#2531Fixes#2593
Signed-off-by: Gleb Natapov <gleb@scylladb.com>
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20170713162014.15343-1-duarte@scylladb.com>
(cherry picked from commit b8235f2e88)
DowngradingConsistencyRetryPolicy uses live replicas count from
Unavailable exception to adjust CL for retry, but when there are pending
nodes CL is increased internally by a coordinator and that may prevent
retried query from succeeding. Adjust live replica count in case of
pending node presence so that retried query will be able to proceed.
Fixes#2535
Message-Id: <20170710085238.GY2324@scylladb.com>
(cherry picked from commit 739dd878e3)
Use name of the existing preceeding column with restriction
(last_column) instead of assuming that the column right after the
current column already has restrictions.
This will yield an error message that is different from that of
Cassandra, albeit still a correct one.
Fixes#2421
Signed-off-by: Botond Dénes <bdenes@scylladb.com>
Message-Id: <40335768a2c8bd6c911b881c27e9ea55745c442e.1499781685.git.bdenes@scylladb.com>
(cherry picked from commit 33bc62a9cf)
With this change, we ask all the shard to handle the ranges provided by
user and we use selective_token_range_sharder to split the ranges and
ignore the ranges do not belong to the current shard.
(cherry picked from commit b10e961a64)
Conflicts:
repair/repair.cc
It is like ring_position_range_sharder but it works with
dht::token_range. This sharder will return the ranges belong to a
selected shard.
(cherry picked from commit d835cf2748)
Repair today has a semaphore limiting the number of ongoing checksum
comparisons running in parallel (on one shard) to 100. We needed this
number to be fairly high, because a "checksum comparison" can involve
high latency operations - namely, sending an RPC request to another node
in a remote DC and waiting for it to calculate a checksum there, and while
waiting for a response we need to proceed calculating checksums in parallel.
But as a consequence, in the current code, we can end up with as many as
100 fibers all at the same stage of reading partitions to checksum from
sstables. This requires tons of memory, to hold at least 128K of buffer
(even more with read-ahead) for each of these fibers, plus partition data
for each. But doing 100 reads in parallel is pointless - one (or very few)
should be enough.
So this patch adds another semaphore to limit the number of checksum
*calculations* (including the read and checksum calculation) on each shard
to just 2. There may still be 100 ongoing checksum *comparisons*, in
other stages of the comparisons (sending the checksum requests to other
and waiting for them to return), but only 2 will ever be in the stage of
reading from disk and checksumming them.
The limit of 2 checksum calculations (per shard) applies on the repair
slave, not just to the master: The slave may receive many checksum
requests in parallel, but will only actually work on 2 at a time.
Because the parallelism=100 now rate-limits operations which use very little
memory, in the future we can safely increase it even more, to support
situations where the disk is very fast but the link between nodes has
very high latency.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20170703151329.25716-1-nyh@scylladb.com>
(cherry picked from commit d177ec05cb)
When peer nodes have the same partition data, i.e., with the same
checksum, we currently choose to stream from any of them randomly.
To improve streaming performance, select the peer within the same DC.
This patch is supposed to improve repair perforamnce with multiple DC.
Message-Id: <c6a345b6e8ed2b59f485e53c865241e463b44507.1498490831.git.asias@scylladb.com>
(cherry picked from commit cc02a62756)
Currently, shard zero is the coordinator of the repair. All the work of
checksuming of the local node and sending of the repair checksum rpc
verb is done on shard zero only. This causes other shards being
underutilized.
With this patch, we split the ranges need to be repaired into at least
smp::count ranges, so sizeof(ranges) / smp::count will be assigned to
each shard. For exmaple, we have 8 shards and 256 ragnes, each shard
will repair 32 ranges. Each shard will repair the 32 ranges
sequencially. There will be at most 8 (smp::count) ranges of repair in
parallel.
(cherry picked from commit 47345078ec)
Conflicts:
repair/repair.cc
In "repair: Use more stream_plan" (commit 2043ffc064), we
switched to do stream while doing checksum instead of do stream only
after checksum pahse is completed. We take a parallelism_semaphore
before we do checksum, if there are more than sub_ranges_to_stream
(1024) ranges, we start a stream_plan and wait for the streaming to
complete (still under the parallelism_semaphore). So at most
parallelism_semaphore (100) stream_plans can be in parallel.
The parallelism_semaphore limits the parallelism of both checksum and the
streaming plan. However, it is not necessary to have the same
parallelism for both checksum and streaming, because 1) a streaming
operation itself runs in parallel (handling ranges on all shards in
prallel, sending mutaitons in parallel) , 2) and with more streaming plan
(in worse case 100) means we can write to 100 memtables at the same time
and flush 100 memtables to disk at the same time which can take a lot of
memory.
With this patch, we only allow one stream plan in flight.
(cherry picked from commit 54831a344c)
It's been linked with various performance issues, either by causing
them or making them worse. One example is #1634, and also recently
I have investigated continuous performance degradation that was also
linked to defrag on idle activity.
Until we can figure out how to reduce its impact, we should disable it.
Signed-off-by: Glauber Costa <glauber@glauber.scylladb>
Message-Id: <20170627201109.10775-1-glauber@scylladb.com>
(cherry picked from commit f3742d1e38)
So that they are not left on disk even though we did a clean shutdown.
First part of the fix is to ensure that closed segments are recognized
as not allocating (_closed flag). Not doing this prevents them from
being collected by discard_unused_segments(). Second part is to
actually call discard_unused_segments() on shutdown after all segments
were shut down, so that those whose position are cleared can be
removed.
Fixes#2550.
Message-Id: <1499358825-17855-1-git-send-email-tgrabiec@scylladb.com>
(cherry picked from commit 6555a2f50b)
when do_for_each is in its last iteration and with_semaphore defers
because there's an ongoing cleanup, sstable object will be used after
freed because it was taken by ref and the container it lives in was
destroyed prematurely.
Let's fix it with a do_with, also making code nicer.
Fixes#2537.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170630035324.19881-1-raphaelsc@scylladb.com>
(cherry picked from commit b9d0645199)
This patch is based on 6c8b5fc. It moves the check whether a dropped
type is still used by other types or tables from schema_tables to
the drop_type_statement, as delaying this check to after applying the
mutations can leave the keyspace in a broken state.
Fixes#2490
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <1497466736-28841-1-git-send-email-duarte@scylladb.com>
If we do two truncates in a row, the second will have neither memtable
nor sstable data. Thus we will not write/remove sstables, and thus
get no resulting truncation replay position.
Fixes#2489
Message-Id: <1497378469-6063-1-git-send-email-calle@scylladb.com>
(cherry picked from commit 525730e135)
end_bound() returns temporary object (end_bound_ref), so it cannot be
taken by reference here and used later. Copy instead.
Message-Id: <20170612132328.GJ21915@scylladb.com>
(cherry picked from commit 21197981a)
Fixes#2482
Test should
a.) Wait for the flush semaphore
b.) Only compare segement sets between start and end, not start,
end and inbetwen. I.e. the test sort of assumed we started
with < 2 (or so) segments. Not always the case (timing)
Message-Id: <1496828317-14375-1-git-send-email-calle@scylladb.com>
(cherry picked from commit 0c598e5645)
It is safe to copy column_mapping accros shards. Such guarantee comes at
the cost of performance.
This patch makes commitlog_entry_writer use IDL generated writer to
serialise commitlog_entry so that column_mapping is not copied. This
also simplifies commitlog_entry itself.
Performance difference tested with:
perf_simple_query -c4 --write --duration 60
(medians)
before after diff
write 79434.35 89247.54 +12.3%
(cherry picked from commit 374c8a56ac)
Also: Fixes#2468.
When using member name in an idetifer of generated class or method
idl compiler should strip the trailing '()'.
(cherry picked from commit 4df4994b71)
(part of #2468)
"This series switches repair to use more stream plans to stream the mismatched
sub ranges and use a range generator to produce sub ranges.
Test shows no huge memory is used for repair with large data set.
In addition, we now have a progress reporter in the log how many ranges are processed.
Jun 06 14:18:22 [shard 0] repair - Repair 512 out of 529 ranges, id=1, keyspace=myks, cf=mytable, range=(8526136029525195375, 8549482295083869942]
Jun 06 14:19:55 [shard 0] repair - Repair 513 out of 529 ranges, id=1, keyspace=myks, cf=mytable, range=(8526136029525195375, 8549482295083869942]
Fixes #2430."
* tag 'asias/fix-repair-2430-branch-master-v1' of github.com:cloudius-systems/seastar-dev:
repair: Remove unused sub_ranges_max
repair: Reduce parallelism in repair_ranges
repair: Tweak the log a bit
repair: Use more stream_plan
repair: iterator over subranges instead of list
(cherry picked from commit 419ad9d6cb)
"This series fixes some issues with the thrift_server, namely
ensuring that streams and sockets are properly closed.
Fixes#499Fixes#2437"
* 'thrift-server-fixes/v1' of github.com:duarten/scylla:
thrift/server: Close connections when stopping server
thrift/server: Move connection class to header
thrift/server: Shutdown connection
thrift/server: Close output_stream when connection is done
(cherry picked from commit a6dc21615b)
After change in boot, read_filter is called by distributed loader,
so its update to _filter_file_size is lost. The load variant
which receives foreign components that must do it. We were also
not updating it for newly created sstables.
Fixes#2449.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170606151129.5477-1-raphaelsc@scylladb.com>
(cherry picked from commit 0ca1e5cca3)
At least on Debian8, mk-build-deps -i silently finishes with return code 0
even it fails to install dependencies.
To prevent this, we should manually install the metapackage generated by
mk-build-deps using gdebi.
Fixes#2445
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1496737502-10737-2-git-send-email-syuu@scylladb.com>
(cherry picked from commit a4c392c113)
Installing openjdk-8-jre-headless from jessie-backports breaks texlive on
jessie main repo.
It causes 'Unmet build dependencies' error when building gdb package.
To prevent this, force insatlling texlive from jessie-backports before start
building gdb.
Fixes#2444
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1496737502-10737-1-git-send-email-syuu@scylladb.com>
(cherry picked from commit 5608842e96)
sstable::data_size() is used by rebuild_statistics() which only
returns uncompressed data size, and the function called by it
expects actual disk space used by all components.
Boot uses add_sstable() which correctly updates the stat with
sstable::bytes_on_disk(). That's what needs to be used by
r__s() too.
Fixes#1592
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170525210055.6391-1-raphaelsc@scylladb.com>
(cherry picked from commit 3b5ad23532)
"fixes a problem in which memory requirement for loading in-memory
components of sstables is very high due to unlimited parallelism."
* 'mem_requirement_sstable_load_v2_2' of github.com:raphaelsc/scylla:
database: fix indentation of distributed_loader::open_sstable
database: reduce memory requirement to load sstables
sstables: loads components for a sstable in parallel
sstables: enable read ahead for read of in-memory components
sstables: make random_access_reader work with read ahead
(cherry picked from commit ef428d008c)
partial sstable files aren't being removed after each failed attempt
to flush memtable, which happens periodically. If the cause of the
failure is ENOSPC, memtable flush will be attempted forever, and
as a result, column family may be left with a huge amount of partial
files which will overwhelm subsequent boot when removing temporary
TOC. In the past, it led to OOM because removal of temporary TOC
took place in parallel.
Fixes#2407.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170525015455.23776-1-raphaelsc@scylladb.com>
(cherry picked from commit b7e1575ad4)
Streaming ususally takes long time to complete. Abort it on false
positive idle detection can be very wasteful.
Increase the abort timeout from 10 minutes to a very large timeout, 300
minutes. The real idle session will be aborted eventually if other
mechanisms, e.g., streaming manager has gossip callback for on_remove
and on_restart event to abort, do not abort the session.
Fixes#2197
Message-Id: <57f81bfebfdc6f42164de5a84733097c001b394e.1494552921.git.asias@scylladb.com>
(cherry picked from commit f792c78c96)
When compacting a fully expired sstable, we're not allowing that sstable
to be purged because expired cell is *unconditionally* converted into a
dead cell. Why not check if the expired cell can be purged instead using
gc before and max purgeable timestamp?
Currently, we need two compactions to get rid of a fully expired sstable
which cells could have always been purged.
look at this sstable with expired cell:
{
"partition" : {
"key" : [ "2" ],
"position" : 0
},
"rows" : [
{
"type" : "row",
"position" : 120,
"liveness_info" : { "tstamp" : "2017-04-09T17:07:12.702597Z",
"ttl" : 20, "expires_at" : "2017-04-09T17:07:32Z", "expired" : true },
"cells" : [
{ "name" : "country", "value" : "1" },
]
now this sstable data after first compaction:
[shard 0] compaction - Compacted 1 sstables to [...]. 120 bytes to 79
(~65% of original) in 229ms = 0.000328997MB/s.
{
...
"rows" : [
{
"type" : "row",
"position" : 79,
"cells" : [
{ "name" : "country", "deletion_info" :
{ "local_delete_time" : "2017-04-09T17:07:12Z" },
"tstamp" : "2017-04-09T17:07:12.702597Z"
},
]
now another compaction will actually get rid of data:
compaction - Compacted 1 sstables to []. 79 bytes to 0 (~0% of original)
in 1ms = 0MB/s. ~2 total partitions merged to 0
NOTE:
It's a waste of time to wait for second compaction because the expired
cell could have been purged at first compaction because it satisfied
gc_before and max purgeable timestamp.
Fixes#2249, #2253
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170413001049.9663-1-raphaelsc@scylladb.com>
(cherry picked from commit a6f8f4fe24)
_underlying is created with _range, which is captured by
reference. But range_and_underlyig_reader is moved after being
constructed by do_with(), so _range reference is invalidated.
Fixes#2377.
Message-Id: <1494492025-18091-1-git-send-email-tgrabiec@scylladb.com>
(cherry picked from commit 0351ab8bc6)
The code that removes each sstable runs in a thread. Parallel
removing of a lot of sstables may start a lot of threads each of which
is taking 128k for its stack. There is no much benefit in running
deletion in parallel anyway, so fix it by deleting sstables sequentially.
Fixes#2384
Message-Id: <20170516103018.GQ3874@scylladb.com>
(cherry picked from commit c7ad3b9959)
mutation_partition has a slicing constructor which is supposed to copy
only the rows from the query range. The rows are located using
nonwrapping_range::lower_bound() and
nonwrapping_range::lower_bound(). Those two have two different
implementations chosen with SFINAE. One is using std::lower_bound(),
and one is using container's built in lower_bound() should it
exist. We're using intrusive tree in mutation_partition, so
container's lower_bound() is preferred. It's O(log N) whereas
std::lower_bound() is O(N), because tree's iterator is not random
access.
However, the current rule for picking container's lower_bound() never
triggers, because lower_bound() has two overloads in the container:
./range.hh:618:14: error: decltype cannot resolve address of overloaded function
typename = decltype(&std::remove_reference<Range>::type::upper_bound)>
^~~~~~~~
As a result, the overload which uses std::lower_bound() is used.
Spotted when running perf_fast_forward with wide partition limit in
cache lifted off. It's so slow that I timeouted waiting for the result
(> 16 min).
Fixes#2395.
Message-Id: <1495048614-9913-1-git-send-email-tgrabiec@scylladb.com>
(cherry picked from commit 3fc1703ccf)
It can't make the leap from dht::ring_position to
stdx::optional<range_bound<dht::ring_position>> for some reason.
(cherry picked from commit ba31619594)
> tls: make shutdown/close do "clean" handshake shutdown in background
> tls: Make sink/source (i.e. streams) first class channel owners
> native-stack: Make sink/source (i.e. streams) first class channel owners
More close() fixes, pointed out by Tomek.
From http://github.com/avikivity/scylla exponential-sharder/v3.
The sharder, which takes a range of tokens and splits it among shards, is
slow with large shard count and the default
murmur3_partitioner_ignore_msb_bits.
This patchset fixes excessive iteration in sstable sharding metadata writer and
nonsignular range scans.
Without this patchset, sealing a memtable takes > 60 ms on a 48-shard
system. With the patchset, it drops below the latency tracker threshold I
used (5 ms).
Fixes#2392.
(cherry picked from commit 84648f73ef)
intersection() returns an optional range with the intersection of the
this range and the other, specified range.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
(cherry picked from commit 1f9359efba)
"This series adds private repository support to scylla-housekeeping"
* 'amnon/housekeeping_private_repo_v3' of github.com:cloudius-systems/seastar-dev:
scylla-housekeeping service: Support private repositories
scylla-housekeeping-upstart: Use repository id, when checking for version
scylla-housekeeping: support private repositories
(cherry picked from commit eb69fe78a4)
The "_specs" array contains column specifications that have the bind
marker name if there is one. That results in
get_partition_key_bind_indices() not being able to look up a column
definition for such columns. Fix the issue by keeping track of the
actual column specifications passed to add() like Cassandra does.
Fixes#2369
(cherry picked from commit a45e656efb4c6478d80e4dfc18de99b94712eeba)
Fix the CQL front-end to populate the partition key bind index array in
result message prepared metadata, which is needed for CQL binary
protocol v4 to function correctly.
Fixes#2355.
(cherry picked from commit ebd76617276e660c590cec0a07e97e82422111df)
Tested-by: Shlomi Livne <shlomi@scylladb.com>
Message-Id: <1494257274-1189-1-git-send-email-penberg@scylladb.com>
"This series fixes the user after free issue in gossip and elimates the
duplicated / unnecessary mark alive operations.
Fixes#2341"
* tag 'asias/gossip_fix_mark_alive/v1' of github.com:cloudius-systems/seastar-dev:
gossip: Ignore callbacks and mark alive operation in shadow round
gossip: Ingore the duplicated mark alive operation
gossip: Fix user after free in mark_alive
(cherry picked from commit 1e04731fa0)
"The logic responsible for converting counter updates to counter shards was
not covered by unit tests and didn't transform counter cells inside static
rows.
This series fixes the problem and makes sure that the tests cover both
static rows and transformation logic.
Fixes #2334."
* tag 'pdziepak/static-counter-updates-1.7/v1' of github.com:cloudius-systems/seastar-dev:
tests/counter: test transform_counter_updates_to_shards
tests/counter: test static columns
counters: transform static rows from updates to shards
We estimate number of partitions for a given range of a column familiy
and split the range into sub ranges contains fewer partitions as a
checksum unit.
The estimation is wrong, because we need to count the partitions on all
the shards, instead of only counting the local shard.
Fixes#2299
Message-Id: <7876285bd26cfaf65563d6e03ec541626814118a.1493817339.git.asias@scylladb.com>
(cherry picked from commit 66e3b73b9c)
During the changes in the way the housekeeping check for newer version
and warn about it in the installation the UUID part was removed but kept
in the sarounding if.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <20170426075724.7132-1-amnon@scylladb.com>
(cherry picked from commit b59c95359d)
We set the scheduler wakeup granularity to 500usec, because that is the
difference in runtime we want to see from a waking task before it
preempts the running task (which will usually be Scylla). Scheduling
other processes less often is usually good for Scylla, but in this case,
one of the "other processes" is also a Scylla thread, the one we have
been using for marking ticks after we have abandoned signals.
However, there is an artifact from the Linux scheduler that causes those
preemption to be missed if the wakeup granularity is exactly twice as
small as the sched_latency. Our sched_latency is set to 1ms, which
represents the maximum time period in which we will run all runnable
tasks.
We want to keep the sched_latency at 1ms, so we will reduce the wakeup
granularity so to something slightly lower than 500usec, to make sure
that such artifact won't affect the scheduler calculations. 499.99usec
will do - according to my tests, but we will reduce it to a round
number.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <20170427135039.8350-1-glauber@scylladb.com>
(cherry picked from commit 14b9aa2285)
Currently, fully expired sstable[1] is unconditionally chosen for compaction
by DTCS, but that may lead to a compaction loop under certain conditions.
Let's consider that an almost expired sstable is compacted, and it's not
deleted yet, and that the new sstable becomes expired before its ancestor is
deleted.
Because this new sstable is expired, it will be chosen by DTCS, but it will
not be purged because 'compacted undeleted' sstables are taken into account
by calculation of max purgeable timestamp and prevents expired data from
being purged. The problem is that this sequence of events can keep happening
forever as reported by issue #2260.
NOTE: This problem was easier to reproduce before improvement on compaction
of expired cells, because fully expired sstable was being converted into a
sstable full of tombstones, which is also considered fully expired.
Fixes#2260.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170428233554.13744-1-raphaelsc@scylladb.com>
(cherry picked from commit 687a4bb0c2)
"Fixes #2327.
Fixes #2326."
* 'tgrabiec/fix-promoted-index-parsing-1.7' of github.com:cloudius-systems/seastar-dev:
sstables: Fix incorrect parsing of cell names in promoted index
sstables: Fix find_disk_ranges() to not miss relevant range tombstones
Attempting to create huge zones may introduce significant latency. This
patch introduces the maximum allowed zone size so that the time spent
trying to allocate and initialising zone is bounded.
Fixes#2335.
Message-Id: <20170428145916.28093-1-pdziepak@scylladb.com>
(cherry picked from commit f5cf86484e)
Range tombstones are serialized to cell names in this place:
_sst.maybe_flush_pi_block(_out, start, {});
Note that the column set is empty. This is correct. A range tombstone
only has a clustering part. The cell name is deserialized by promoted
index reader using mp_row_consumer::column, like this:
mp_row_consumer::column col(schema, std::move(col_name),
api::max_timestamp); return std::move(col.clustering);
The problem is, column constructor assumes that there is always a
component corresponding to a cell name if the table is not dense, and
will pop it from the set of components (the clustering field):
, cell(!schema.is_dense() ? pop_back(clustering) : (*(schema.regular_begin())).name())
promoted index block which starts or ends with a range tombstone will
appear as having incorrect bounds. This may result in an incorrect
value for data file range start to be calculated.
Fixes#2327.
Suppose the promoted index looks like this:
block0: start=1 end=2
block1: start=4 end=5
start and end are cell names of the first and last cell in the block.
If there is a range tombstone covering [2,3], it will be only in
block0, because it is no longer in effect when block1 starts. However,
slicing the index for [3, +inf], which intersects with the tombstone,
will yield block1. That's because the slicing looks for a block with
an end which is greater than or equal to the start of the slice:
if (!found_range_start) {
if (!range_start || cmp(range_start->value(), end_ck) <= 0) {
range_start_pos = ie.position() + offset;
We should take into account that any given block may actually contain
information for anything up to the start of the next block, so instead
of using end_ck, effectively use next block's start_ck (exclusive).
Fixes#2326.
"Currently eviction is performed until occupancy of the whole region
drops below the 85% threshold. This may take a while if region had
high occupancy and is large. We could improve the situation by only
evicting until occupancy of the sparsest segment drops below the
threshold, as is done by this change.
I tested this using a c-s read workload in which the condition
triggers in the cache region, with 1G per shard:
lsa-timing - Reclamation cycle took 12.934 us.
lsa-timing - Reclamation cycle took 47.771 us.
lsa-timing - Reclamation cycle took 125.946 us.
lsa-timing - Reclamation cycle took 144356 us.
lsa-timing - Reclamation cycle took 655.765 us.
lsa-timing - Reclamation cycle took 693.418 us.
lsa-timing - Reclamation cycle took 509.869 us.
lsa-timing - Reclamation cycle took 1139.15 us.
The 144ms pause is when large eviction is necessary.
Statistics for reclamation pauses for a read workload over
larger-than-memory data set:
Before:
avg = 865.796362
stdev = 10253.498038
min = 93.891000
max = 264078.000000
sum = 574022.988000
samples = 663
After:
avg = 513.685650
stdev = 275.270157
min = 212.286000
max = 1089.670000
sum = 340573.586000
samples = 663
Refs #1634."
* tag 'tgrabiec/lsa-reduce-reclaim-latency-v3' of github.com:cloudius-systems/seastar-dev:
lsa: Reduce reclamation latency
tests: Add test for log_histogram
log_histogram: Allow non-power-of-two minimum values
lsa: Use regular compaction threshold in on-idle compaction
tests: row_cache_test: Induce update failure more reliably
lsa: Add getter for region's eviction function
(cherry picked from commit fccbf2c51f)
[avi: adjustments for 1.7's heap vs. master's log_histogram]
Problem is that column family field of task wasn't being set for resharding,
so column family wasn't being properly removed from compaction manager.
In addition to fixing this issue, we'll also interrupt ongoing compactions
when dropping a column family, exactly like we do with shutdown.
Fixes#2291.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170418125807.7712-1-raphaelsc@scylladb.com>
(cherry picked from commit e78db43b79)
streaming generates lots of small sstables with large token range,
which triggers O(N^2) in space in interval map.
level 0 sstables will now be stored in a structure that has O(N)
in space complexity and which will be included for every read.
Fixes#2287.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170417185509.6633-1-raphaelsc@scylladb.com>
(cherry picked from commit 11b74050a1)
We take a reference of endpoint_state entry in endpoint_state_map. We
access it again after code which defers, the reference can be invalid
after the defer if someone deletes the entry during the defer.
Fix this by checking take the reference again after the defering code.
I also audited the code to remove unsafe reference to endpoint_state_map entry
as much as possible.
Fixes the following SIGSEGV:
Core was generated by `/usr/bin/scylla --log-to-syslog 1 --log-to-stdout
0 --default-log-level info --'.
Program terminated with signal SIGSEGV, Segmentation fault.
(this=<optimized out>) at /usr/include/c++/5/bits/stl_pair.h:127
127 in /usr/include/c++/5/bits/stl_pair.h
[Current thread is 1 (Thread 0x7f1448f39bc0 (LWP 107308))]
Fixes#2271
Message-Id: <529ec8ede6da884e844bc81d408b93044610afd2.1491960061.git.asias@scylladb.com>
(cherry picked from commit d27b47595b)
The previous fix removed the additional insertion of "min rp" per source
shard based on whether we had processed existing CF:s or not (i.e. if
a CF does not exist as sstable at all, we must tag it as zero-rp, and
make whole shard for it start at same zero.
This is bad in itself, because it can cause data loss. It does not cause
crashing however. But it did uncover another, old old lingering bug,
namely the commitlog reader initiating its stream wrongly when reading
from an actual offset (i.e. not processing the whole file).
We opened the file stream from the file offset, then tried
to read the file header and magic number from there -> boom, error.
Also, rp-to-file mapping was potentially suboptimal due to using
bucket iterator instead of actual range.
I.e. three fixes:
* Reinstate min position guarding for unencoutered CF:s
* Fix stream creating in CL reader
* Fix segment map iterator use.
v2:
* Fix typo
Message-Id: <1490611637-12220-1-git-send-email-calle@scylladb.com>
(cherry picked from commit b12b65db92)
Discarding blocks on large RAID volume takes too much time, user may suspects
the script doesn't works correctly, so it's better to skip, do discard directly on each volume instead.
Fixes#1896
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1489533460-30127-1-git-send-email-syuu@scylladb.com>
(cherry picked from commit b65d58e90e)
Fixes#2173
Per-shard min positions can be unset if we never collected any
sstable/truncation info for it, yet replay segments of that id.
Wrap the lookups to handle "missing data -> default", which should have been
there in the first place.
Message-Id: <1490185101-12482-1-git-send-email-calle@scylladb.com>
(cherry picked from commit c3a510a08d)
This patch exposes Scylla's Prometheus port by default. You can now use
the Scylla Monitoring project with the Docker image:
https://github.com/scylladb/scylla-grafana-monitoring
To configure the IP addresses, use the 'docker inspect' command to
determine Scylla's IP address (assuming your running container is called
'some-scylla'):
docker inspect --format='{{ .NetworkSettings.IPAddress }}' some-scylla
and then use that IP address in the prometheus/scylla_servers.yml
configuration file.
Fixes#1827
Message-Id: <1490008357-19627-1-git-send-email-penberg@scylladb.com>
(cherry picked from commit 85a127bc78)
On Ubuntu 14.04, the lsblk doesn't have '-p' option, but
`scylla_setup` try to get block list by `lsblk -pnr` and
trigger error.
Current simple pattern will match all help content, it might
match wrong options.
scylla-test@amos-ubuntu-1404:~$ lsblk --help | grep -e -p
-m, --perms output info about permissions
-P, --pairs use key="value" output format
Let's use strict pattern to only match option at the head. Example:
scylla-test@amos-ubuntu-1404:~$ lsblk --help | grep -e '^\s*-D'
-D, --discard print discard capabilities
Signed-off-by: Amos Kong <amos@scylladb.com>
Message-Id: <4f0f318353a43664e27da8a66855f5831457f061.1489712867.git.amos@scylladb.com>
(cherry picked from commit 468df7dd5f)
We're cleaning up sstables in parallel. That means cleanup may need
almost twice the disk space used by all sstables being cleaned up,
if almost all sstables need cleanup and every one will discard an
insignificant portion of its whole data.
Given that cleanup is frequently issued when node is running out of
disk space, we should serialize cleanups in every shard to decrease
the disk space requirement.
Fixes#192.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20170317022911.10306-1-raphaelsc@scylladb.com>
(cherry picked from commit 7deeffc953)
"The test allocates objects in batches (allocation is always under a reclaim
lock) of ~3MiB and assumes that it will always succeed because if we cross the
low water mark for free memory (20MiB) in seastar, reclamation will be
performed between the batches, asynchronously.
Unfortunately that's prevented by can_allocate_more_memory(), which fails
segment allocation when we're below the low water mark. LSA currently doesn't
allow allocating below the low water mark.
The solution which is employed across the code base is to use allocating_section,
so use it here as well.
Exposed by recent consistent failures on branch-1.7."
* 'tgrabiec/fix-lsa-async-eviction-test' of github.com:cloudius-systems/seastar-dev:
tests: lsa_async_eviction_test: Allocate objects under allocating section
lsa: Allow adjusting reserves in allocating_section
(cherry picked from commit 434a4fee28)
Fixes#2098
Replay previously did all segments in parallel on shard 0, which
caused heavy memory load. To reduce this and spread footprint
across shards, instead do X segments per shard, sequential per shard.
v2:
* Fixed whitespace errors
Message-Id: <1489503382-830-1-git-send-email-calle@scylladb.com>
(cherry picked from commit 078589c508)
Metrics name should be unique per type.
requests_blocked_memory was registered twice, one as a gauge and one as
derived.
This is not allowed.
Fixes#2165
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <20170314162826.25521-1-amnon@scylladb.com>
(cherry picked from commit 0a2eba1b94)
Metrics should have their unique name. This patch changes
throttled_writes of the queu lenght to current_throttled_writes.
Without it, metrics will be reported twice under the same name, which
may cause errors in the prometheus server.
This could be related to scylladb/seastar#250
Fixes#2163.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <20170314081456.6392-1-amnon@scylladb.com>
(cherry picked from commit 295a981c61)
The current test for whether or not the filesystem is mounted is weak
and will fail if multiple pieces of the hierarchy are mounted.
util-linux ships with a mountpoint command that does exactly that,
so we'll use that instead.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <1488742801-4907-1-git-send-email-glauber@scylladb.com>
(cherry picked from commit 2d620a25fb)
By default behavior is kept the same. There are deployments in which we
would like to mount data and commitlog to different places - as much as
we have avoided this up until this moment.
One example is EC2, where users may want to have the commitlog mounted
in the SSD drives for faster writes but keep the data in larger, less
expensive and durable EBS volumes.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <1488258215-2592-1-git-send-email-glauber@scylladb.com>
(cherry picked from commit 9e61a73654)
We have:
auto halves = range.split(midpoint, dht::token_comparator());
We saw a case where midpoint == range.start, as a result, range.split
will assert becasue the range.start is marked non-inclusive, so the
midpoint doesn't appear to be contain()ed in the range - hence the
assertion failure.
Fixes#2148
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Signed-off-by: Asias He <asias@scylladb.com>
Message-Id: <93af2697637c28fbca261ddfb8375a790824df65.1489023933.git.asias@scylladb.com>
(cherry picked from commit 39d2e59e7e)
"If a node is bootstrapped with auto_boostrap disabled, it will not
wait for schema sync before creating global keyspaces for auth and
tracing. When such schema changes are then reconciled with schema on
other nodes, they may overwrite changes made by the user before the
node was started, because they will have higher timestamp.
To prevent that, let's use minimum timestamp so that default schema
always looses with manual modifications. This is what Cassandra does.
Fixes #2129."
* tag 'tgrabiec/prevent-keyspace-metadata-loss-v1' of github.com:scylladb/seastar-dev:
db: Create default auth and tracing keyspaces using lowest timestamp
migration_manager: Append actual keyspace mutations with schema notifications
(cherry picked from commit 6db6d25f66)
The skip() implementation for the compressed file input stream incorrectly
handled the case of skipping to the end of file: In that case we just need
to update the file pointer, but not skip anywhere in the compressed disk
file; In particular, we must NOT call locate() to find the relevant on-disk
compressed chunk, because there is none - locate() can only be called on
actual positions of bytes, not on the one-past-end-of-file position.
Fixes#2143
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20170308100057.23316-1-nyh@scylladb.com>
(cherry picked from commit 506e074ba4)
db_apply() expects to be given a time point at which the request will
time out. Originally, do_apply_counter_update() passed 0, which meant
that all requests were timed out if do_apply() needed to wait. The
caller of do_apply_counter_update() is already given a correct timeout
time point so the only thing needed to fix this problem it to propagate
it properly inside do_apply_counter_update() to the call to do_apply().
Fixes#2119.
Message-Id: <20170307104405.5843-1-pdziepak@scylladb.com>
logalloc::reclaim_lock prevents reclaim from running which may cause
regular allocation to fail although there is enough of free memory.
To solve that there is an allocation_section which acquire reclaim_lock
and if allocation fails it run reclaimer outside of a lock and retries
the allocation. The patch make use of allocation_section instead of
direct use of reclaim_lock in memtable code.
Fixes#2138.
Message-Id: <20170306160050.GC5902@scylladb.com>
(cherry picked from commit d7bdf16a16)
If query_time is time_point::min(), which is used by
to_data_query_result(), the result of subtraction of
gc_grace_seconds() from query_time will overflow.
I don't think this bug would currently have user-perceivable
effects. This affects which tombstones are dropped, but in case of
to_data_query_result() uses, tombstones are not present in the final
data query result, and mutation_partition::do_compact() takes
tombstones into consideration while compacting before expiring them.
Fixes the following UBSAN report:
/usr/include/c++/5.3.1/chrono:399:55: runtime error: signed integer overflow: -2147483648 - 604800 cannot be represented in type 'int'
Message-Id: <1488385429-14276-1-git-send-email-tgrabiec@scylladb.com>
(cherry picked from commit 4b6e77e97e)
Since gcc-5/stretch=5.4.1-2 removed from apt repository, we nolonger able to
build gcc-5.
To avoid dead link, use launchpad.net archives instead of using apt-get source.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1488189378-5607-1-git-send-email-syuu@scylladb.com>
(cherry picked from commit ba323e2074)
Fixes the following UBSAN warning:
core/semaphore.hh:293:74: runtime error: reference binding to misaligned address 0x0000006c55d7 for type 'struct basic_semaphore', which requires 8 byte alignment
Since the field was not initialied properly, probably also fixes some
user-visible bug.
Message-Id: <1488368222-32009-1-git-send-email-tgrabiec@scylladb.com>
(cherry picked from commit 0c84f00b16)
Failing to close a file properly before destroying file's object causes
crashes.
[tgrabiec: fixed typo]
Fixes#2122.
Message-Id: <20170221144858.GG11471@scylladb.com>
(cherry picked from commit 0977f4fdf8)
Set murmur3_partitioner_ignore_msb_bits to 12 (enabling the new sharding
algorithm), but do this in scylla.yaml rather than the built-in defaults.
This avoids changing the configuration for existing clusters, as their
scylla.yaml file will not be updated during the upgrade.
Message-Id: <20170214123253.3933-1-avi@scylladb.com>
(cherry picked from commit 9b113ffd3e)
"This series contains some fixes and a unit test for the logic responsible
for locking counter cells."
* 'pdziepak/cell-locking-fixes/v1' of github.com:cloudius-systems/seastar-dev:
tests: add test for counter cell locker
cell_locking: fix schema upgrades
cell_locker: make locker non-movable
cell_locking: allow to be included by anyone
(cherry picked from commit b8c4b35b57)
ninja-build-1.6.0-2.fc23.src.rpm on fedora web site deleted for some
reason, but there is ninja-build-1.7.2-2 on EPEL, so we don't need to
backport from Fedora anymore.
Fixes#2087
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1487155729-13257-1-git-send-email-syuu@scylladb.com>
(cherry picked from commit 9c8515eeed)
"This series makes sure that schemas containing both counter and non-counter
regular or static columns are not allowed."
* 'pdziepak/disallow-mixed-schemas/v1' of github.com:cloudius-systems/seastar-dev:
schema: verify that there are no both counter and non-counter columns
test/mutation_source: specify whether to generate counter mutations
tests/canonical_mutation: don't try to upgrade incompatible schemas
(cherry picked from commit 9e4ae0763d)
Merge commit 45b6070832 used butchered version of storage_proxy
patch to adjust to rpc timer change instead the one I've sent. This
patch fixes the differences.
Message-Id: <20170206095237.GA7691@scylladb.com>
(cherry picked from commit 3c372525ed)
scylla-housekeeping requires to run 'restart mode' for check the version during
scylla-server restart, which wasn't called on systemd timer so added it.
Existing scylla-housekeeping.timer renamed to scylla-housekeeping-daily.timer,
since it is running 'daily mode'.
Fixes#1953
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1486180031-18093-1-git-send-email-syuu@scylladb.com>
(cherry picked from commit e82932b774)
The comparator constructor took schema by value instead of const l-ref
and, consequently, later tried to access object that has been destroyed
long time ago.
Message-Id: <20170202135853.8190-1-pdziepak@scylladb.com>
(cherry picked from commit 37b0c71f1d)
"Before, the logic for releasing writes blocked on dirty worked like this:
1) When region group size changes and it is not under pressure and there
are some requests blocked, then schedule request releasing task
2) request releasing task, if no pressure, runs one request and if there are
still blocked requests, schedules next request releasing task
If requests don't change the size of the region group, then either some request
executes or there is a request releasing task scheduled. The amount of scheduled
tasks is at most 1, there is a single releasing thread.
However, if requests themselves would change the size of the group, then each
such change would schedule yet another request releasing thread, growing the task
queue size by one.
The group size can also change when memory is reclaimed from the groups (e.g.
when contains sparse segments). Compaction may start many request releasing
threads due to group size updates.
Such behavior is detrimental for performance and stability if there are a lot
of blocked requests. This can happen on 1.5 even with modest concurrency
because timed out requests stay in the queue. This is less likely on 1.6 where
they are dropped from the queue.
The releasing of tasks may start to dominate over other processes in the
system. When the amount of scheduled tasks reaches 1000, polling stops and
server becomes unresponsive until all of the released requests are done, which
is either when they start to block on dirty memory again or run out of blocked
requests. It may take a while to reach pressure condition after memtable flush
if it brings virtual dirty much below the threshold, which is currently the
case for workloads with overwrites producing sparse regions.
I saw this happening in a write workload from issue #2021 where the number of
request releasing threads grew into thousands.
Fix by ensuring there is at most one request releasing thread at a time. There
will be one releasing fiber per region group which is woken up when pressure is
lifted. It executes blocked requests until pressure occurs."
* tag 'tgrabiec/lsa-single-threaded-releasing-v2' of github.com:cloudius-systems/seastar-dev:
tests: lsa: Add test for reclaimer starting and stopping
tests: lsa: Add request releasing stress test
lsa: Avoid avalanche releasing of requests
lsa: Move definitions to .cc
lsa: Simplify hard pressure notification management
lsa: Do not start or stop reclaiming on hard pressure
tests: lsa: Adjust to take into account that reclaimers are run synchronously
lsa: Document and annotate reclaimer notification callbacks
tests: lsa: Use with_timeout() in quiesce()
(cherry picked from commit 7a00dd6985)
"This series introduces support for counters. The implementation of
counters more or less follows the design described on our wiki page [1].
Counter cells contain many shards with replicas being able to modify
and announce new versions only of the shards that they own. Historically,
there were three types of shards: local, remote and global. In these
patches only support for the global ones is added.
[1] https://github.com/scylladb/scylla/wiki/Counters
Currently, counters are only enabled as experimental features as there
still several things that need to be done before they become production
ready. Namely, the performance is expected to be quite poor (especially
for writes), there is no proper tracing support and timed out counter
requests may not be recognized and dropped early. There are also no
counter-related metrics.
However, apart from these problems there are no other missing parts of
counter implementation and they are expected to work correctly.
Fixes #577."
* 'pdziepak/counters/v3-rebased' of github.com:cloudius-systems/seastar-dev: (38 commits)
perf_simple_query: add counter tables tests
thrift: add support for counter operations
cql3: allow counters in CREATE TABLE statements
cql3: selection: do not panic when seeing counters
storage_proxy: support counter updates
storage_proxy: add get_live_endpoints()
cql3: add counter increment and decrement operations
db: add operations for applying counter updates
counters: implement transforming counter deltas to shards
add infrastructure for locking counter cells
add fnv1a hasher
position_in_partition: add feed_hash()
position_in_partition: add functions for querying object type
types: make counter_type_impl report its cql3_type
transport: encode counters as long_type
mutation_partition: make for_each_cell() accessible outside source file
messaging_service: add COUNTER_MUTATION verb
storage_service: add COUNTERS feature
idl: add idl description of consistency level
schema: make is_counter() return correct value
...
The leader receives counter updates as deltas which have to be
transformed to counter shards. In order to do that, current local shard
of the modified counter cell needs to be read, logical clock incremented
and the value modified by the specified delta.
The leader receives counter update in a form of deltas which need to be
transformed to counter shards. In order to do that the node needs to
read its current state of the modified counter cells. Since this is
essentially a read-modify-write opertation an appropriate locking
mechanism is needed.
Counter cell locker introduced in this patch uses a hashtable of
partition entry each containing a hashtable of cell entries. Inside a
cell entry there is a semaphore used for synchronization. Once no longer
needed cell entries and partition entries are removed.
In order to avoid deadlocks cell entries are always locked in the same
order which is the lexicographical order of (clustering key, column id)
pairs. Note that schema changes are not a difficulty since they do not
make it possible to change ordering of such pairs.
Support for deletion of counters is limited in a way that once deleted
they cannot be used again (i.e. tombstone always wins, regardless of the
timestamp). Logic responsible for merging two counter cells already
makes sure that tombstones are handled properly, but it is also
necessary to ensure that higher level tombstones always cover counters.
Live counter cells are collections of shards, each one representing the
sum of all operations performed by a particular replica. This commits
introduces an in-memory representation of counters as well as basic
operations such as merge, difference and hashing.
* seastar c1dbd89...f07f8ed (3):
> Merge "Introduce when_all_succeed()" from Paweł
> tests: adjust collectd test for metric API change
> Merge "DNS query support" from Calle
* seastar 397685c...c1dbd89 (13):
> lowres_clock: drop cache-line alignment for _timer
> net/packet: add missing include
> Merge "Adding histogram and description support" from Amnon
> reactor: Fix the error: cannot bind 'std::unique_ptr' lvalue to 'std::unique_ptr&&'
> Set the option '--server' of tests/tcp_sctp_client to be required
> core/memory: Remove superfluous assignment
> core/memory: Remove dead code
> core/reactor: Use logger instead of cerr
> fix inverted logic in overprovision parameter
> rpc: fix timeout checking condition
> rpc: use lowres_clock instead of high resolution one
> semaphore: make semaphore's clock configurable
> rpc: detect timedout outgoing packets earlier
Includes treewide change to accomodate rpc changing its timeout clock
to lowres_clock.
Includes fixup from Amnon:
collectd api should use the metrics getters
As part of a preperation of the change in the metrics layer, this change
the way the collectd api uses the metrics value to use the getters
instead of calling the member directly.
This will be important when the internal implementation will changed
from union to variant.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1485457657-17634-1-git-send-email-amnon@scylladb.com>
need_preempt() is always true in debug mode. Because of that, this loop
will never be executed. Rewrite it as a do-while loop so we are sure
that it is executed at least once - or exactly once in debug mode.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <1485913079-1283-1-git-send-email-glauber@scylladb.com>
The per-node limit will be total memory divided by number of shards
instead of just total memory. For example, when Scylla is started with
-c16 -m16G, the commit log will induce flushes on given shard when
unflushed data exceeds on that shard 62MB instead of 1GB.
Fixes#2046.
Message-Id: <1485874534-10939-1-git-send-email-tgrabiec@scylladb.com>
Currently the test does not wait for cache update
to finish before carrying on with the checks.
This makes the test nondeterministic and purely wrong
because checks expect update to be finished.
This patch changes the test to wait for update to finish.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Message-Id: <2a99bba24b1628466d3495332b48ef3ccdb43c26.1485862389.git.piotr@scylladb.com>
An earlier patch of mine was using should_yield to do the same. That
is a better direction, but should_yield() was demonstrably more
expensive so for now we'll go with need_preempt() - since this is
hurting pretty much every latency-dependent workload.
I am also including the scripts that I have used to measure and
compare the various versions of this patch.
The data() method already returns a bytes_opt so there's no need to call to_bytes_opt() again.
Fixes compliation failure on CentOS:
In file included from ./cql3/query_options.hh:51:0,
from ./cql3/cql_statement.hh:47,
from ./cql3/statements/raw/select_statement.hh:45,
from build/release/gen/cql3/CqlParser.hpp:65,
from build/release/gen/cql3/CqlParser.cpp:44:
./cql3/values.hh: In function 'bytes_opt to_bytes_opt(const cql3::raw_value&)':
./cql3/values.hh:184:37: error: no matching function for call to 'to_bytes_opt(bytes_opt)'
return to_bytes_opt(value.data());
Message-Id: <1485761863-28236-1-git-send-email-penberg@scylladb.com>
The reason is the same as why foreground writes are reported instead of
total writes (049ae37d08): It is much easier to see what is going on
this way.
Also fixes a typo in a counter's description.
Fixes#1217
Message-Id: <20170129093412.GS11469@scylladb.com>
This patch series adds support for "unset values" that were introduced
in CQL binary protocol v4. They allow bound statements to skip updates
to some or all of the bound variables.
Unset values are specified using the BoundStatement.unset() method in
the Java driver:
http://docs.datastax.com/en/drivers/java/3.1/com/datastax/driver/core/BoundStatement.html#unset-int-
and using the UNSET_VALUE constant in the Python driver:
https://datastax.github.io/python-driver/api/cassandra/query.html#cassandra.query.UNSET_VALUEFixes#2039.
* 'penberg/cql-unset-values/v2' of github.com:cloudius-systems/seastar-dev:
transport/server: CQL unset value support
cql3/statements/select_statement: Unset value support
cql3/user_types: Unset value support
cql3/tuples: Unset value support
cql3/maps: Unset value support
cql3/sets: Unset value support
cql3/lists: Unset value support
cql3/constants: UNSET_VALUE constant
cql3/constants: Unset value support
cql3/attributes: Unset value support
types.hh: Add field_name_as_string() to user_type_impl type
cql3: Introduce raw_value and raw_value_view types
The quota check is quite old at the moment, and dates back to a time in
which the infrastructure in seastar threads was lacking a lot. It is a
bad check since it will not take into consideration the size of the
partition or the time it takes to merge them.
A better check would at least take need_preempt() into account, so that
we would respect the task quota. That check is now embedded into
should_yield(), so there would no need to check anything else.
Although should_yield() does the job, it is still currently quite
expensive. And because we are in a seastar thread with a computationally
intensive loop, it can hurt latency a lot.
So as a temporary measure, let's at least check for need_preempt() - as
it is hurting real users at the moment - and soon work on making
should_yield() cheaper.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
update is one of our biggest sources of performance issues as far as the
cache is concerned. systemtap can be useful in helping tracking some of
them down.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Currently, the code is using bytes_opt and bytes_view_opt to represent
CQL values, which can hold a value or null. In preparation for
supporting a third state, unset value introduced in CQL v4, introduce
new raw_value and raw_value_view types and use them instead.
The new types are based on boost::variant<> and are capable of holding
null, unset values, and blobs that represent a value.
This reverts commit d61002cc33.
Introduced a regression in row_cache_alloc_stress.
The problem is that reclaim_from_evictable() evicts way too much after
the refactor due to the stop condition not taking into account how
much data was evicted so far and only looking at occupancy of the
minimal segment. This may lead to eviction of the whole region.
parse_time() adds hourse, minutes, etc to a final value 'result'.
However, it is of type std::chrono::nanoseconds which means it is not
zeroed at initialization unless it is explicitly asked to do so.
Fixed debug mode failures in types_tyes and cql_query_test.
Message-Id: <20170125155239.1253-1-pdziepak@scylladb.com>
"This patchset properly implements range_tombstone_list::difference(),
which was very broken. We add unit tests for the function and ensure
we always randomly generate range_tombstones in other unit tests so
other problems aren't hidden."
This patch ensures the mutation_merger emits any deferred tombstones
that it still may be holding before closing the stream.
Together with the range_tombstone_list: Properly implement
difference() patch set, this fixes breakage of streamed_mutation_test
and row_cache_test.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20170123195643.9876-1-duarte@scylladb.com>
* seastar 6d80c6a...397685c (4):
> Merge "add label to the io_queue" from Amnon
> rpc: Modify the shutdown code to wait and handle exceptions
> tls.cc: Fix shutdown_input/output to conform with expected socket behaviour
> core: Add counter for polls
The difference method wasn't properly implemented. The version in this
patch correctly computes the difference and returns a range tombstone
list contains those range tombstones in "this" but absent from the
other, specified range tombstone list.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
* seastar 38aaa4a...6d80c6a (2):
> DPDK: Change the metrics registration with label support
> metric: Fix the error: could not convert {...} from <brace-enclosed initializer list> to struct metric_definition_impl
Add a counter field to RELEASE, just before the date, and fix it at zero.
This allows custom package builds to override it in a way that sorts before
the official packages.
Example:
Official release: 1.6.0-0.20160120.<githash>
Custom release 1: 1.6.0-1.avi.20160121.<githash>
Custom release 2: 1.6.0-2.avi.20160122.<githash>
The counter (0/1/2) ensures that the build number dominates over the date
when sorting.
Message-Id: <20170122102814.19649-1-avi@scylladb.com>
From Avi:
In many cases, batch statements are used to mutate a single partition, or
a number of partitions that is smaller than the number of statements within
the batch. We can detect this case and reduce the numbers of mutations
applied, and in some cases, convert a logged batch into an unlogged batch.
Ref #1689.
continuous_data_consumer::fast_forward_to() returns a future which was
later ignored by data_consume_context::fast_forward_to().
With the current implementation, the future in question is always ready
and that's why the problem didn't manifest itself in the form of crashes
or invalid results.
Message-Id: <20170120105746.7300-1-pdziepak@scylladb.com>
Batch statements are often used to insert multiple rows into the same
partition. Recognize this case and merge mutations to the same partition.
If the result is a single mutation, there is an additional win (already
present in the code), where a logged batch can be converted into an unlogged
batch.
Ref #1689.
Add a boolean to short circuit the read path on empty range
hoping for some speedup.
tested in read write with cs using:
cl=QUORUM duration=1m -mode native cql3 -rate threads=700 -node localhost
Will do some additional benchmark.
Fixes#1056
Signed-off-by: Benoît Canet <benoit@scylladb.com>
Message-Id: <20170118194451.16836-1-benoit@scylladb.com>
That's because a single shard is used to calculate generation for new
sstables in upload directory, and that will result in that single shard
sharing all the resources with other shards.
For refresh without upload dir, it currently works fine because we
reshuffle column family dir instead.
flush_upload_dir() is now a free function, takes a distributed database
object, and uses calculate_shard_from_sstable_generation() to decide
which shard will move sstable using its own generation namespace.
Fixes#2008.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <b0cccf7bbb61416ff8718bac92fdca90cc5fb9c9.1484253232.git.raphaelsc@scylladb.com>
This patch ensures that when adding a shared sstable, we select only
one cpu to update that column family's stats. This is important so we
don't overestimated the on-disk size of sstables when resharding
This fixes only a temporary miscount of the current load, since shared
sstables are eventually re-written, but a fixes a permanent miscount
of the total load.
Refs #1592
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20170119144823.31041-1-duarte@scylladb.com>
Currently eviction is performed until occupancy of the whole region
drops below the 85% threshold. This may take a while if region had
high occupancy and is large. We could improve the situation by only
evicting until occupancy of the sparsest segment drops below the
threshold, as is done by this change.
I tested this using a c-s read workload in which the condition
triggers in the cache region, with 1G per shard:
lsa-timing - Reclamation cycle took 12.934 us.
lsa-timing - Reclamation cycle took 47.771 us.
lsa-timing - Reclamation cycle took 125.946 us.
lsa-timing - Reclamation cycle took 144356 us.
lsa-timing - Reclamation cycle took 655.765 us.
lsa-timing - Reclamation cycle took 693.418 us.
lsa-timing - Reclamation cycle took 509.869 us.
lsa-timing - Reclamation cycle took 1139.15 us.
The 144ms pause is when large eviction is necessary.
The change improves worst case latency. Reclamation time statistics
over 30 second period after cache fills up, in microseconds:
Before:
avg = 1524.283148
stdev = 11021.021118
min = 12.934000
max = 144356.000000
sum = 257603.852000
samples = 169
After:
avg = 1317.362414
stdev = 1913.542802
min = 263.935000
max = 19244.600000
sum = 175209.201000
samples = 133
Refs #1634.
Message-Id: <1484730859-11969-1-git-send-email-tgrabiec@scylladb.com>
* seastar 240b0bf...ff098c8 (15):
> metrics::impl::shard(): check if reactor is initialized before using it
> reactor: introduce engine_is_ready()
> fix metric name
> Merge "Add label support to the metric layer" from Amnon
> core: Avoid memory leak when submission to syscall_work_queue fails
> core: Avoid memory leak when submission to smp_message_queue fails
> core: append_challenged_posix_file_impl: Make exception-safe
> Merge "Log backtrace in report_failed_future" from Tomasz
> install-dependencies.sh: add systemtap-sdt-dev to Ubuntu/Debian dependencies
> core: add fsqual.cc/.hh to core
> dpdk: Fix compile error with rte_pci.h
> fstream_test: fix spurious failures due to BOOST_REQUIRE_EQUAL thread-unsafety
> reactor: unregister metrics of queue on shard 0
> build: track system header changes too
> Prometheus: do not rely on collectd for the hostname
close() operation is like a destructor, it cannot fail. It just
reports errors, but close itself succeeds. So we should proceed with
the closing even if it fails.
Message-Id: <1484245886-7269-1-git-send-email-tgrabiec@scylladb.com>
As Tomek pointed out, previous code, regardless of version mismatch, of generating
comparator description string was not correct (as in: in sync with origin).
This modifies it to look at
1.) Actual clustring size
2.) Compound-ness
3.) Dense-ness
to determine whether we should generate a compound desc, and whether it
should contain a trailing utf8-desc type.
v2: Simplify non-dense base column addition and ensure it handles
thrift non-utf8 (as per comments from tomek)
Message-Id: <1484670171-18362-1-git-send-email-calle@scylladb.com>
As the metrics migration progressed, some include to scollectd.hh left
behind.
Because of the nature of the scollecd implementation those include
brings alot of code with them to the header files and eventually to many
source file.
This patch remove those include and add a missing include to
storage_proxy.cc.
The reason the compiler didn't complain is an indication to the
problematic nature of those include in the first place.
Before this patch, change in metrics.hh would cause 169 files to
compile, after this change 17.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1484667536-2185-1-git-send-email-amnon@scylladb.com>
Fixes#2019
According to the Java driver and cassandra, all versions < 3
include the PK in the comparator descriptor string.
This broke for us when bumping the cassandra version 2.1 -> 2.2
Message-Id: <1484657580-14411-1-git-send-email-calle@scylladb.com>
This patch ensures that the host only announces and registers the
MATERIALIZED_VIEWS feature if it was started with the experimental
flag.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20170116123412.21365-1-duarte@scylladb.com>
After we call unlink_leftmost_without_rebalance(), we must unlink all
elements before mutatation is destroyed. We did this properly from
~reader, but it would not be called if reader construction failed,
which it may.
Message-Id: <1484572581-6537-1-git-send-email-tgrabiec@scylladb.com>
Since ce083308a1
"random_mutation_generator: Generate RTs by default" random mutation
generator produces range tombstones. However, so far the tests were run
with all features disabled (because of incomplete initialization of all
services) which meant that RANGE_TOMBSTONE feature was not enabled and
the code couldn't handle range tombstones that weren't just prefixes.
This patch solves the problem by forcing all features to be enabled when
tests are run.
Message-Id: <20170116103324.22956-1-pdziepak@scylladb.com>
It still has problems:
- while resharding a very large leveled compaction strategy table, a huge
amount of tiny sstables are generated, overwhelming the file descriptor
limits
- there is a large impact on read latency while resharding is going on
(cherry picked from commit cf27d44412)
(forward-ported from branch-1.6)
Since scylla-housekeeping running as scylla user, it doesn't have a permission
to create a file on /etc/scylla.d.
So introduce /var/lib/scylla-housekeeping which owns by scylla user, place uuid
file on the directory.
Fixes#2009
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1484235946-12463-1-git-send-email-syuu@scylladb.com>
query_mutations_locally() takes one_or_two_partition_ranges by
reference and requires, indirectly, that it is kept alive until
operation resolves. However, we were passing expiring value to it, the
result of unwrap().
Fixes dtest failure in consistent_bootstrap_test.py:TestBootstrapConsistency.consistent_reads_after_bootstrap_test
Another potential problem was that we were dereferencing "s" in the same
expression which move-constructs an argument out of it.
Message-Id: <1484222759-4967-1-git-send-email-tgrabiec@scylladb.com>
On DPDK 16.11 dpdk_nic_bind.py is renamed to dpdk-devbind.py, so we are
getting "file not found" both on packaging and scripts, fixed that.
Also fixed inconsistent packaging.
Since Seastar copied dpdk_nic_bind.py to its scripts/ directory, there're two
different versions of the script, .rpm/.deb packaging different one:
dist/redhat: seastar/dpdk/tools/dpdk_nic_bind.py
dist/ubuntu: seastar/scripts/dpdk_nic_bind.py
That's won't work because we sharing setup scripts between two
distributions, so I changed dist/ubuntu package to use DPDK one.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1484191955-28006-1-git-send-email-syuu@scylladb.com>
Explicitly generate tables' IDs of tables from the system_traces KS using
generate_legacy_id() in order to ensure all Nodes create these tables with
the same IDs.
This is going to prevent hitting issue #420.
Fixes#1976
Signed-off-by: Vlad Zolotarov <vladz@scylladb.com>
Message-Id: <1484153725-31030-1-git-send-email-vladz@scylladb.com>
This patch changes the random_mutation_generator so it generates range
tombstones by default. This fixes a bug where reversibly applying
range tombstones wasn't being tested.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20170110164822.28747-1-duarte@scylladb.com>
boost::intrusive::value_traits_pointers was introduced in boost 1.56, while
we also support boost 1.55. Replace with an equivalent expression.
(with additions by Asias)
Message-Id: <20170110084700.19994-1-avi@scylladb.com>
Commit f0c28e1 ("db/schema_tables: Add schema_functions and
schema_aggregates tables") forgot to add the newly added tables to the
db::schema_tables::ALL list, which is used for authorization checks, for
example.
Fixes the following auth_test.py dtest failures:
('Unable to connect to any servers', {'127.0.0.1': Unauthorized('Error from server: code=2100 [Unauthorized] message="User cathy has no SELECT permission on <table system.schema_functions> or any of its parents"',)})
Message-Id: <1484045277-4997-1-git-send-email-penberg@scylladb.com>
"Intended to reduce memory usage when resharding by sharing sstable
components among shards. File descriptors are also shared from now
on, meaning that a much smaller number of file descriptors will be
used during resharding.
Fixes #1951."
branch 'excessive_memory_usage_v4' of github.com:raphaelsc/scylla
* 'excessive_memory_usage_v4' of github.com:raphaelsc/scylla:
db: avoid excessive memory usage during resharding
checked_file_impl: add support to dup
sstables: group sstable components that can be shared among shards
sstables: rename sstable member
After resharding, sstables may be owned by all shards, which
means that file descriptors and memory usage for metadata will
increase by a factor equal to number of shards. That can easily
lead to OOM.
SSTable components are immutable, so they can be stored in one
shard and shared with others that need it. We use the following
formula to decide which shard will open the sstable and share
it with the others: (generation % smp::count), which is the
inverse of how we calculate generation for new sstables.
So if no resharding is performed, everything is shard-local.
With this approach, resource usage due to loaded sstables will
be evenly distributed among shards.
For this approach to work, we now only populate keyspaces from
shard 0. It's now the sole responsible for iterating through
column family dirs. In addition, most of population functions
are now free and take distributed database object as parameter.
Fixes#1951.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
"This patch series adds support for CQL 3.3.1. The changes to CQL are listed
here:
https://github.com/apache/cassandra/blob/cassandra-2.2/doc/cql3/CQL.textile#changes
The following CQL features are already supported by Scylla:
- TRUNCATE TABLE alias
- Double-dollar string literals
- Aggregate functions: MIN, MAX, SUM, and AVG
This series adds the following CQL features:
- New data types: tinyint, smallint, date, and time
- CQL binary protocol v4 (required by the new data types)
- Advertise Cassandra 2.2.8 version from Scylla so that drivers correctly
detect the presence of CQL 3.3.1
The following CQL features are not supported by Scylla:
- Role-based access control (issue #1941)
- JSON data type
- User-defined functions (UDFs)
- User-defined aggregates (UDAs)
The following CQL binary protocol v4 changes are not implemented by this
series:
- Read_failure and Write_failure error codes are not implemented.
They error codes not used by the smart drivers but as they are
propagated to application code, we eventually need to wire them up
to our storage proxy implementation.
- Function_failure error code is only used by user-defined functions
and the fromJson function, which are not implemented by Scylla.
Fixes #1284."
* 'penberg/cql-3.3.1/v5' of github.com:cloudius-systems/seastar-dev:
version: Bump Cassandra version to 2.2.8
db/schema_tables: Add schema_functions and schema_aggregates tables
tests/type_tests: TIME type test cases
tests/cql_query_test: TIME type test cases
cql3: TIME data type support
tests/type_tests: DATE type test cases
tests/cql_query_test: DATE type test cases
cql3: DATE type support
date.h: 64-bit year and days representation
licenses: Add utils/date.h license
utils/date.h: Import date and time library sources
tests/type_tests: TINYINT and SMALLINT type test cases
tests/cql_query_test: TINYINT and SMALLINT type test cases
cql3: TINYINT and SMALLINT data type support
types: Fix integer_type_impl::parse_int() for bytes
mutation_result_merger::get() assumes that the merged result may be a
short read if at least one of the partial results is a short read (in
other words, if none of the partial results is a short read, then the
merged result is also not a short read). However this is not true;
because we update the memory accounter incrementally, we may stop
scanning early. All the partial results are full; but we did not scan
the entire range.
Fix by changing the short_read variable initialization from `no`
(which assumes we'll encounter a short read indication when processing
one of the batches) to `this->short_read()`, which also takes into
account the memory accounter.
Fixes#2001.
Message-Id: <20170108111315.17877-1-avi@scylladb.com>
We need 64-bit year and days representation to support the boundary
values of the CQL data type, which is implemented using Joda Time
library's DateTime type.
Noone overrides file_writer::write() so there is no reason to inhibit
optimisations and cause compiler to emit indirect calls.
Message-Id: <20170104163618.26251-1-pdziepak@scylladb.com>
Ubuntu Packaging Guide says if there's no upstream package (means it's not
ported from Debian), revision should be "0ubuntu1", not "ubuntu1" which is we
currently using.
On Debian, Debian Policy Manual says it's conventional to restart revision from 1 when upstream version increased, so we should specify it to "1".
To do it in single script, we will generate the revision on building time.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1483498658-27491-1-git-send-email-syuu@scylladb.com>
This patch gets housekeeping to create a uuid file if a path to a uuid
file is upplied but the file is missing.
Because it import the uuid lib, uuid parameters where renamed.
Fixes#1987
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1483866553-13855-2-git-send-email-amnon@scylladb.com>
We intend to share immutable sstable components among shards to
reduce excessive memory usage when resharding shared sstables.
This change is about grouping those components into a structure,
and using foreign ptr to make sure that the structure will be
deleted by whichever shard created it.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Rename _components to _recognized_components because _components
will be used to name a field with shareable components.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
* seastar 1c8e389...240b0bf (15):
> file/dup: don't decrease refcnt twice when file is explicitly closed
> reactor: Add missing CentOS 7.2 dependency systemtap-sdt-devel
> reactor: Cleaning the smp queue metrics when shuting down
> metrics: metrics keep the value map while unregistering
> change the reactor load metrics to utilization
> Merge "ASan fiber switches" from Paweł
> tls: Add missing credentials_builder::set_client_auth method
> collectd: create metrics with the right format
> io_queue: remove owner number from metric name
> reactor: change the load metric name to load
> Merge "reactor: stop using signals for task_quota timer"
> metrics: Allow initializing the metric_group in its constructor
> Update DPDK to 16.11
> Revert "rpc: Avoid using zero-copy interface of output_stream"
> core::metrics_groups: add a clear() method
Transform the supervisor_notify() and related functions into
the "supervisor" class and place this class implementation in
a separate .cc file.
This is going to fix the compilation breakage of tests introduced
by a
commit 8014adc2a1
init: serialize the creation of system_traces KS objects
Signed-off-by: Vlad Zolotarov <vladz@scylladb.com>
Message-Id: <1483663955-20096-1-git-send-email-vladz@scylladb.com>
"Reduce the size of mutation_partition by implementing intrusive set using
bi::rbtree_algorithms directly and using tree nodes optimized for size.
This will reduce the size of mutation_partition by:
24 bytes + <number of cql rows> * 8 bytes
This should have a positive impact on performance because mutation_partitions
are stored both in memtable and cache.
Fixes #742."
* 'haaawk/742' of github.com:cloudius-systems/seastar-dev:
intrusive_set: rename size() to calculate_size()
Make intrusive_set_external_comparator::_value_traits static
Implement intrusive set using rbtree_algorithms
mutation_partition: make apply_reversibly_intrusive_set nongeneric
mutation_partition: take schema in find_row and clustered_row
mutation_partition: Extract intrusive set logic to a class.
mutation_partition: Replace value_comp with key_comp calls
Before this patch system table writes were not writing to commit log
because database::add_column_family() disables writes to commit log
for the table which is added if _commitlog is not set at that
time. Fix by initializing commit log before system tables are created.
Fixes#1986.
Fixes recent regression in
batch_test.py:TestBatch.replay_after_schema_change_test after
scylla-jmx was updated to not flush system tables on nodetool flush.
Could cause system keyspace writes to be delayed for more than before
under heavy write workload. Refs #1926.
Message-Id: <1483618117-4535-1-git-send-email-tgrabiec@scylladb.com>
During a range scan, we try to avoid sorting according to partition range
when we can do so. This is when we scan fewer than smp::count shards --
each shard's range is strictly ordered with respect to the others.
However, we use the wrong key for the sort -- we use the shard number. But
if we started at shard s > 0 and wrapped around to shard 0, then shard 0's
range will be after the range belonging to shard s, but will sort before it.
Fix by storing the iteration order as the sort key. We use that when we
know that shards do not overlap (shards < smp::count) and the index within
the source partition range vector when they do.
Fixes#1998.
Message-Id: <20170105114253.17492-1-avi@scylladb.com>
Serialize the creation of a system_traces KS objects when
they do not exist - the initial cluster boot.
Avoid creating them in parallel by different cluster Nodes
in order to avoid issue #420.
Signed-off-by: Vlad Zolotarov <vladz@scylladb.com>
Message-Id: <1483552503-12873-3-git-send-email-vladz@scylladb.com>
This hopefully will make it more apparent that
the time complexity of this method is O(N) not O(1).
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
_value_traits can be shared among all instances
and there's no need to store it in every single one.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
1.6 truncates paged queries early to avoid overrunning server memory
with too-large query results, but in the case of partition range queries,
this terminates too early due to an uninitialized variable holding the
maximum result size. This results in slow performance due to additional
round trips.
Fix by initializing the maximum result size from the result_memory_tracker
running on the coordinating shard.
Fixes#1995.
Message-Id: <20170105103915.10633-1-avi@scylladb.com>
This new implementation takes less memory because it
does not store comparator.
It also uses tree nodes optimized for size. This means
that instead of storing an enum field |color| they embed
this information inside pointer to parent.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
apply_reversibly_intrusive_set is used only in one place
and always with rows_type. There's no need for it to be generic.
This will allow changing intrusive set implementation.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
The integer_type_impl::parse_int() function uses boost::lexical_cast()
under the hood, which parses 8-bit numbers as characters. Fix the
function to lexical cast to 64-bit integer and convert the result to
integer_type_impl template type.
Previously, if the Prometheus port (by default, 0.0.0.0:9180) could not
be opened, the following message appeared in the log about 10 seconds into
the run, and Scylla crashed.
ERROR 2017-01-01 19:31:04,066 [shard 0] seastar - Exiting on unhandled exception: std::system_error (error system:98, Address already in use)
The puzzled user would have no idea *which* address was already in use, why,
or why Scylla stopped.
In this patch, before the above message we get the much more informative
message:
ERROR 2017-01-01 19:58:19,080 [shard 0] init - Could not start Prometheus API server on 0.0.0.0:9180: std::system_error (error system:98, Address already in use)
We continue to print the original message - and exit - in this case,
under the assumption that it's better not to run the database while
improperly configured.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20170102121304.2060-1-nyh@scylladb.com>
The test case is interested in the upper boundary of 32-bit integer
because we already test the lower boundary in assertions below. The old
test passed, of course, but it wasn't very interesting.
Message-Id: <1483522773-6008-1-git-send-email-penberg@scylladb.com>
Fix the following build breakage:
FAILED: build/release/gen/cql3/CqlParser.o
g++ -MMD -MT build/release/gen/cql3/CqlParser.o -MF build/release/gen/cql3/CqlParser.o.d -std=gnu++1y -g -Wall -Werror -fvisibility=hidden -pthread -I/home/penberg/scylla/seastar -I/home/penberg/scylla/seastar/fmt -I/home/penberg/scylla/seastar/build/release/gen -march=nehalem -Ifmt -DBOOST_TEST_DYN_LINK -Wno-overloaded-virtual -DFMT_HEADER_ONLY -DHAVE_HWLOC -DHAVE_NUMA -DHAVE_LZ4_COMPRESS_DEFAULT -O2 -DBOOST_TEST_DYN_LINK -Wno-maybe-uninitialized -DHAVE_LIBSYSTEMD=1 -I. -I build/release/gen -I seastar -I seastar/build/release/gen -c -o build/release/gen/cql3/CqlParser.o build/release/gen/cql3/CqlParser.cpp
In file included from ./query-request.hh:31:0,
from ./locator/token_metadata.hh:51,
from ./locator/abstract_replication_strategy.hh:29,
from ./database.hh:26,
from ./service/storage_proxy.hh:44,
from ./db/schema_tables.hh:43,
from ./db/system_keyspace.hh:46,
from ./cql3/functions/function_name.hh:45,
from ./cql3/selection/selectable.hh:48,
from ./cql3/selection/writetime_or_ttl.hh:45,
from build/release/gen/cql3/CqlParser.hpp:63,
from build/release/gen/cql3/CqlParser.cpp:44:
./tracing/tracing.hh:357:5: error: ‘scollectd’ does not name a type
scollectd::registrations _registrations;
^~~~~~~~~
Message-Id: <1482939751-8756-1-git-send-email-penberg@scylladb.com>
Commit d41cd48a made the is_joined() method a future<bool> because
only cpu 0 knows its real value. This makes this function inconvenient
to use. So this patch reverts commit d41cd48a, and instead sets this
flag's value on all shards, so each shard can read its value locally
(and immediately).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20161228160450.5831-1-nyh@scylladb.com>
* seastar f32e4c2...1c8e389 (2):
> Merge "migrate network related seastar collectd metrics to the new metrics registration API" from Vlad
> file: add dup() support
This patch adds an overload to the apply() function,
which takes a clustering_row by reference, to copy. This will be
needed by future patches, when merging base table updates with the
existing data.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <1482881106-3202-1-git-send-email-duarte@scylladb.com>
This patch removes references to the old begin_range_tombstone and
end_range_tombstone mutation_fragments, which have been replaced by a
single range_tombstone fragment.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <1482880820-2831-1-git-send-email-duarte@scylladb.com>
- Change an exception type thrown by a tracing::tracing::set_trace_probability()
to make it different from the one thrown by an std::stod() when it fails to
parse a given string.
- Catch the std::out_of_range exception thrown by a tracing::tracing::set_trace_probability() and
wrap the exception string into the httpd::bad_param_exception() object.
- Throw a httpd::bad_param_exception() with a
"Bad format in a probability value: <a user given probability string value>"
message if std::invalid_argument is caught.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Message-Id: <1465300738-1557-1-git-send-email-vladz@cloudius-systems.com>
Refuse to boot if we don't have at least 1 GiB per shard, unless in developer
mode.
The primary violator here is docker, but since it starts in developer mode,
it won't get fixed. We need some extra logic for this case.
Message-Id: <20161221090222.28677-1-avi@scylladb.com>
"This patchset contains fixes for the changes introduced in "Query result
size limiting". It also improves handling of short data reads.
I order to minimise chances of digest mismatch during data queries replicas
that were asked just to return a digest also keep track of the size of the
data (in the IDL representation) so that they would stop at the same point
nodes doing full data queries would. Moreover, data queries are not
affected by per-shard memory limit and the coordinator sends individual
result size limits to replicas in order not to depend on hardcoded values.
It is still possible to get digest mismatches if the IDL changes (e.g. a
new field is added), but, hopefully, that won't be a serious problem."
* 'pdziepak/short-read-fixes/v4' of github.com:cloudius-systems/seastar-dev:
query: introduce result_memory_accounter::foreign_state
storage_proxy: fix short reads in parallel range queries
storage_proxy: pass maximum result size to replicas
mutation_partition: use result limiter for digest reads
query: make result_memory_limiter constants available for linker
result_memory_limiter: add accounter for digest reads
idl: allow writers to use any output stream
result_memory_limiter: split new_read() to new_{data, mutation}_read()
idl: is_short_read() was added in 1.6
mutation_partition: honour allowed_short_read for static rows
storage_proxy: fix _is_short_read computation
storage_proxy: disallow short reads if got no live rows
storage_proxy: don't stop after result with no live rows
This reverts commit aa392810ff, reversing
changes made to a24ff47c637e6a5fd158099b8a65f1191fc2d023; it uses
boost::intrusive::detail directly, which it must not, and doesn't compile on
all boost versions as a consequence.
* seastar 0b98024...f32e4c2 (11):
> Merge "Moving the reactor counters to the metric layer" from Amnon
> metrics: Metrics function should take variable as a refernce
> Revert "Merge ""Moving the reactor counters to the metric layer from Amnon"
> Merge ""Moving the reactor counters to the metric layer from Amnon
> Revert "fstream: Auto-close data_sink and data_source"
> rpc: Avoid resource unit leaks on failure
> fstream: Auto-close data_sink and data_source
> http: Move metrics registration to the metrics layer
> output_stream: add batching to zero copy interface
> Revert "slab: Move the metrics registration to the metrics layer"
> slab: Move the metrics registration to the metrics layer
"Reduce the size of mutation_partition by implementing intrusive set using
bi::rbtree_algorithms directly and using tree nodes optimized for size.
This will reduce the size of mutation_partition by:
24 bytes + <number of cql rows> * 8 bytes
This should have a positive impact on performance because mutation_partitions
are stored both in memtable and cache.
Fixes #742."
* 'haaawk/742' of github.com:cloudius-systems/seastar-dev:
intrusive_set: rename size() to calculate_size()
Make intrusive_set_external_comparator::_value_traits static
Implement intrusive set using rbtree_algorithms
mutation_partition: make apply_reversibly_intrusive_set nongeneric
mutation_partition: take schema in find_row and clustered_row
mutation_partition: Extract intrusive set logic to a class.
mutation_partition: Replace value_comp with key_comp calls
This hopefully will make it more apparent that
the time complexity of this method is O(N) not O(1).
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
_value_traits can be shared among all instances
and there's no need to store it in every single one.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
This new implementation takes less memory because it
does not store comparator.
It also uses tree nodes optimized for size. This means
that instead of storing an enum field |color| they embed
this information inside pointer to parent.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
apply_reversibly_intrusive_set is used only in one place
and always with rows_type. There's no need for it to be generic.
This will allow changing intrusive set implementation.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
file output streams take the responsibility of closing the file, they
will close the file as part of closing the stream.
During sstable writing we create sstable object and keep file
references there as well. Sstable object also has responsibility for
closing the files, and does so from sstable::~sstable().
Double close was supposed to be avoided by a construct like this:
writer.close().get();
_file = {};
However if close() failed, which can happen when write-ahead failed,
_file would not be cleared, and both the writer and sstable would
close the file. This will result in a crash in
append_challenged_posix_file_impl::close(), which is not prepared to
be closed twice.
Another problem is that if exception happened before we reached that
construct, we still should close the writer. Currently we don't, so
there's no double close on the file, but that's a bug which needs to
be fixed and once that's fixed double close on _file will be even more
likely.
The fix employed here is to not keep files inside sstable object when
writing. As soon as the writer is constructed, it's the only owner of
the file.
Fixes#1764.
Message-Id: <1482428648-22553-1-git-send-email-tgrabiec@scylladb.com>
Range queries used to be performed sequentially and the shard performing
part of the read was reading state of the merger's memory accounter
directly. Now, they may be performed in parallel so it is safer to just
pass relevant data by value to the intersted shards so that they are not
reading something that another shard is modyfing at the same time.
Since query is done in parallel there is a chance of overread. However,
the parallelism is high only in sparsely populated tables and that's
when the overread is less serious problem.
Since a1cafed370 "storage_proxy: handle
range scans of sparsely populated tables" nonsingular range queries may
be performed in parallel on multiple shards. The consequence of this
that result may be added to the merger out of order. This requires more
complex logic for handling short reads.
As soon as mutation_result_merger gets a short read it starts to discard
all subsequently received results that are known to contain partitions
with larger keys.
Then when the final result is being prepared the merger may need to
combine and sorts results which ordering is not known. If at least one
of these results is a short one all partitions with larger keys are
removed.
Due to request being performed in parallel it is possible that even
though there was a short read the merger has got enough live data to
satisfy specified limits. If this has happened the short read flag is
not set on the final result.
We may want to change the default individual result size limit in the
future. If it is provided by the coordinator and not hardcoded in the
replicas this can be done without causing data query digest mismatches
or wasteful mutation query results.
Even if we are performing a digest query we should do proper result
memory accounting so that the result ends exactly in the same place that
it would if it was a data query. This is to avoid digest mismatches
between replicas.
It was the observation that ring_position_range_sharder doesn't support
wrapping ranges that started the nonwrapping_range madness, but that
class still has some leftover wrapping ranges. Close the circle by
removing them.
Message-Id: <20161123153113.8944-1-avi@scylladb.com>
Digest reads differ from data reads in a way that they do not really
consume any memory. We still want them to stop in the same place that
data reads would, but the per-shard semaphore shouldn't be updated by
them.
Original IDL generated code was hardcoded to always use bytes_ostream.
This patch makes the output stream a template parameter so that any
valid output stream can be used.
Unfortunately, making IDL writers generic requires updates in the code
that uses them, this is fixed in C++17 which would be able to deduce the
parameter in most cases.
For data queries it is very important that all replicas get limited in
the same place (this includes replicas returning only digest). That's
why they shouldn't be affected by per-shard result memory limit.
Moreover, we should make sure that individual memory limits are the
same, making the coordinator provide it for replicas which allow to
safely change it in the future.
Mutation queries are not as sensitive but it is still beneficial to make
sure that all replicas use the same individual limit.
If after reconciliation the coordinator ends up with no live rows and
short reads are allowed a retry may not make any progress if replicas
end their reads in the same place. The solution is to disallow short
reads on retries which are caused by final result having no live rows.
mutation_result_merger merges results from different shards and stops as
soon as a shard returned a short read or memory usage on the merging
shard is too high. However, it should never stop unless at least one
live rows is in the merged result.
* 'tgrabiec/calculate-hash-once-compaction' of github.com:cloudius-systems/seastar-dev:
sstables: Calculate key hash only once during compaction
tests: sstables: Add more test cases to tombstone_purge_test
db: Expose column_family::add_sstable
tests: sstables: Ensure timestamps are increasing
tests: sstables: Simplify tombstone_purge_test
- moved to seastar thread
- extracted sstable creation and validation logic
- reduced code duplication
- switched to mutation_reader assertions
- used result of compact_sstable() to locate the new sstable
- rather than setting gc timestamp in the past, bump the clock
before compacting
Some code paths were obtaining db_clock timestamp to only convert it
to gc_clock later. Avoid this. In the future we could make gc_clock
cheaper cause it has low precision.
Message-Id: <1482401190-2035-1-git-send-email-tgrabiec@scylladb.com>
Shared sstables will now be resharded in the same order to guarantee
that all shards owning a sstable will agree on its deletion nearly
the same time, therefore, reducing disk space requirement.
That's done by picking which column family to reshard in UUID order,
and each individual column family will reshard its shared sstables
in generation order.
Fixes#1952.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <87ff649ed24590c55c00cbb32bffd8fa2743e36e.1482342754.git.raphaelsc@scylladb.com>
Currently server-side timestamps use a clock with millisecond
precision. Timestamps have microsecond resolution, with lower bits
used to serialize mutations originating from given client.
Timestamps for column drops always use just the millisecond base. A
column drop which is executed after an insert may thus be given lower
timestamp than the insert, even when the two are serialized on the
client side over same connection.
Use microsecond precision to reduce chances of that event.
This is supposed to fix sporadic failures of
schema_test.py:TestSchema.drop_column_queries_test dtest.
Message-Id: <1482343119-27698-1-git-send-email-tgrabiec@scylladb.com>
Refs #1943.
* 'tgrabiec/optimize-bloom-filter' of github.com:cloudius-systems/seastar-dev:
db: Compute key hash once in partition_presence_checker
bloom_filter: Allow checking presence using pre-hashed key
db: Use incremental selector in partition_presence_checker
RPC messaging service is initialized before the Tracing service, so
we should prevent creation of tracing spans before the service is
fully initialized.
We will use an already existing "_down" state and extend it in a way
that !_down equals "started", where "started" is TRUE when the local
service is fully initialized.
We will also split the Tracing service initialization into two parts:
1) Initialize the sharded object.
2) Start the tracing service:
- Create the I/O backend service.
- Enable tracing.
Fixes issue #1939
Signed-off-by: Vlad Zolotarov <vladz@scylladb.com>
Message-Id: <1481836429-28478-1-git-send-email-vladz@scylladb.com>
If a node gets more MUTATION request that it can handle via RPC it will
stop reading from this RPC connection, but this will prevent it from
getting MUTATION_DONE responses for requests it coordinates because
currently MUTATION and MUTATION_DONE messages shares same connection.
To solve this problem this patches moves MUTATION_DONE messages to
separate connection.
Fixes: #1843
Message-Id: <20161201155942.GC11581@scylladb.com>
This reverts commit b81a57e8eb.
With exponential range scanning, we should now be able to survive
msb ignore bits of 12, which allows better sharding on large clusters.
When murmur3_partitioner_ignore_msb_bits = 12 (which we'd like to be the
default), a scan range can be split into a large number of subranges, each
going to a separate shard. With the current implementation, subranges were
queried sequentially, resulting in very long latency when the table was empty
or nearly empty.
Switch to an exponential retry mechanism, where the number of subranges
queried doubles each time, dropping the latency from O(number of subranges)
to O(log(number of subranges)).
If, during an iteration of a retry, we read at most one range
from each shard, then partial results are merged by concatentation. This
optimizes for the dense(r) case, where few partial results are required.
If, during an iteration of a retry, we need more than one range per
shard, then we collapse all of a shard's ranges into just one range,
and merge partial results by sorting decorated keys. This reduces
the number of sstable read creations we need to make, and optimizes for
the sparse table case, where we need many partial results, most of which
are empty.
We don't merge subranges that come from different partition ranges,
because those need to be sorted in request order, not decorated key order.
[tgrabiec: trivial conflicts]
Message-Id: <20161220170532.25173-1-avi@scylladb.com>
This patchset implements the multiple CQL3 statements relating to
materialized views, as well as ensuring other statements now take
materialized views into account. It also adds the necessary internal
data structures to hold materialized view metadata.
This patch adds a set of tests for materialized view schema
handling, complementing the dtests for the same feature.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch introduces the do_with_cql_env_thread() function, which
behaves like do_with_cql_env() except that it executes the
user-specified function in the context of a Seastar thread.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds an utility function that creates a raw select
statement from a set of columns and a where clause. It is intended to
be used to create the prepared select statement used by the view
class.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds the alter_view_statement, which enables users to
change the properties of a materialized view.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds the MATERIALIZED_VIEWS_FEATURE to the set of cluster
features and requires its presence to allow creating a view. This
ensures view schemas can be safely propagated across nodes.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch propagates the for_view argument, used by
statement_restrictions to ensure IS NOT NULL can be used when creating
a materialized view.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch checks for additional permissions when modifying a table
with views, since that update will require reading from the table and
writing into its views.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch forbids dropping a column family if there are still views
associated with it, and also forbids dropping a view through the drop
table statement.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch ensures that changes to a base table's schema
are reflected in that table's materialized views.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
To minimize code duplication, have query_processor use
do_with_parser() instead of manually creating the CqlParser.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds an utility function to rename a column occurring a
textual where clause. It is intended to change a view's where clause
when users alter the underlying base table.
To do this, we rely on functions that transform a textual where clause
into a set of relations, which allows to reliably rename the column.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds an utility function to rename an identifier
occurring in a cql3 relation. This function will be used when renaming
an identifier in a view's where clause.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch builds the mutations to announce a new view. Aside from
including the view schema, we include the base table mutations so
that a node is resilient against receiving create view mutations
before the base table create mutations.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch allows a view schema to be frozen. To unfreeze such a
schema, we add an is_view attribute to the schema idl.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch replaces the add_table_to_schema_mutation() function with
add_table_or_view_to_schema_mutation().
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch extracts update_column_family from schema_tables into
database so it can be used when adding materialized views, in future
patches.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes the drop_column_family() function to remove
a view schema from the list of views of its base table.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds code for parsing the views schema table upon init and
also ensures that when adding a view column family, that we add it to
its base table list of views.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch moves some duplicate code into the
add_column_family_and_create_directory() function. It also saves some
superfluous keyspace lookups and readies the code to be used by
materialized views.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes read_table_mutations() so that it can now
read schemas from other tables besides the column families
schema table.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch ensures we don't provide access to materialized views over
thrift. This includes preventing updates but also omitting them when
describing a keyspace.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds utility functions to keyspace_metadata to select only
the tables or only the views out of all the schemas.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds the view class, which will contains functions related
to populating a view, either from the base table's write path or from
the view building mechanism which copies over already existing data in
the base table.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
The view_ptr class contains a schema_ptr known to represent a
materialized view. It is intended to be used by functions that require
such a schema, and thus obviate the need for the function to check for
schema::is_view().
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds a view_info optional field to the schema. It's
presence indicates the schema represents a materialized view.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch fixes a regression introduced in 0518895, where we counted
one extra row per partition when it contained live, non static rows.
We also simplify the visitor logic further, since now we don't need to
count rows one by one. Also remove a bunch of unused fields.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <1482234083-2447-1-git-send-email-duarte@scylladb.com>
Will allow us to calculate the hash once and use it on many filters
instead of calculating the hash for each filter separately.
Another change made is to avoid precomputing all indexes during filter
operations, and have for_each_index() template instead which invokes a
functor.
This reduces the number of sstables we need to check to only those
whose token range overlaps with the key. Reduces cache update
time. Especially effective with leveled compaction strategy.
Refs #1943.
Incremental selector works with an immutable sstable set, so cache
updates need to be serialized. Otherwise we could mispopulate due to
stale presence information.
Presence checker interface was changed to accept decorated key in
order to gain easy access to the token, which is required by
the incremental selector.
"nonwrapping_range<ring_position> and nonwrapping_range<token> are used
in many places. Let's make an alias for them to make it less verbose.
Also there is a query::partition_range in query-request.hh which is the alias of
nonwrapping_range<ring_position>. query::partition_range is used in
places not related to query at all. Let's unify the usage project wide."
* tag 'asias/repair_dht_token_range/v2' of github.com:cloudius-systems/seastar-dev:
Convert to use dht::partition_range_vector and dht::token_range_vector
dht: Introduce dht::partition_range_vector and dht::token_range_vector
Get rid of query::partition_range
Convert to use dht::partition_range
Convert to use dht::token_range
dht: Rename token_range to token_range_endpoints
dht: Introduce dht::token_range an dht::partition_range
std::vector<dht::partition_range> and std::vector<dht::token_range> are
used in a lot of places, introduce dht::partition_range_vector and
dht::token_range_vector as the alias.
nonwrapping_range<ring_position> and nonwrapping_range<token> are used
in many places. Let's make an alias for them to make it less verbose.
Also there is a query::partition_range in query-request.hh which is the alias of
nonwrapping_range<ring_position>. query::partition_range is used in
places not related to query at all. Let's unify the usage project wide.
"In 7c873f0d (repair: Reduce unnecessary streaming traffic), we optimize
in cases when 1) all the remote nodes has the same checksum and 2) local node
has zero checksum.
In this series, we make the optimization more generec and cover more cases."
* tag 'asias/repair/node_reducer/v3' of github.com:cloudius-systems/seastar-dev:
repair: Reduce unnecessary streaming traffic even more
repair: Add hash specialization for partition_checksum
"This patchset ensures the partition limit is enforced at
the storage_proxy level. Uppers layers like the pager may
already be depending on this behavior."
* 'enforce-row-limit/v3' of https://github.com/duarten/scylla:
query_pagers: Don't trim returned rows
select_statement: Don't always trim result set
query_result_merger: Limit rows
mutation_query: to_data_query_result enforces row limit
"This patchset ensures the partition limit is enforced at
the storage_proxy level. To achieve this, we add the partition
count to query::result, and allow the result_merger to trim
excess partitions."
* 'enforce-partition-limit/v3' of https://github.com/duarten/scylla:
storage_proxy: Decrease limits when retrying command
storage_proxy: Don't fetch superfluous partitions
query::result: Add partition count
column_family: Use counters in query::result::builder
query_result_builder: Use the underlying counters
mutation_partition: Count partitions in query_compacted
mutation_partition: Remove tabs in query_compacted
query::result::builder: Add partition count
query_result_merger: Limit partitions
A case could be made that we should have counters for them no matter
what, since it can help us reason about the distribution of memory among
the groups. But with the hierarchy being broken in 1.5 it becomes even
more important. Now by looking solely at dirty, we will have no idea
about how much memory we are using in those groups.
After this patch, the dirty_memory_manager will register its metrics
for the 3 groups that we have, and the legacy names will be used to show
totals.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <0d04ca4c7e8472097f16a5dc950b77c73766049e.1481831644.git.glauber@scylladb.com>
"Original, naive db::make_streaming_reader() implementation created a set
of memtable and sstable readers for every partition range. This caused
bad interaction with the code limiting sstable readers concurrency and
was suboptimal.
This series introduces multi range mutation reader that takes mutation
source and a sorted, disjoint vector of ranges. It creates only a single
set of memtable and sstable readers and fast forwards it to the next
range once the current one is completed."
* 'pdziepak/multi-range-reader/v1' of github.com:cloudius-systems/seastar-dev:
db: use multi range reader for streaming readers
dht: describe split_range[s]_to_shards() guarantees
repair: remove outdated fixme
test/mutation_reader_test: add multi_range_reader test
tests/mutation_reader: extract key creation code
mutation_reader: add multi_range_reader
A naive approach was to create a set of readers for each range and pass
them all to combining reader. This however performed badly if the number
of ranges was high.
The solution is to use multi range reader which uses only a single set
of readers and fast forwards from range to range when necessary. This
adds another requirement that the ranges passed to
make_streaming_reader() are sorted and disjoint.
We are going to require these functions to return sorted and disjoint
ranges. They already do so (provided that the input ranges are sorted
and disjoint), but if the guarantee is not explicitly stated it may
disappear some day.
So far, the only way to combine outputs of multiple readers was to use
combining reader. It is very general and, in particular, supports case
when the readers emit mutations from overlapping ranges.
However, we have cases (e.g. streaming) when we need to read from
several disjoint ranges. Combining reader is a suboptimal solution as it
requires to creating a reader for each range and ignores the fact that
they do not overlap.
This patch introduces multi_range_mutation_reader which takes a
mutation_source and a sorted set of disjoint ranges. Internally, it uses
mutation_reader::fast_forward_to() to move to the next range once the
current one is completed.
Since storage_proxy::query() now respects the read_command limits, we
can remove the trimming logic from query_pagers.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Trimming the result set is only needed when the query contains an "IN"
relation, an ORDER BY clause, and defines a limit, which is the case
where we query different ranges concurrently. We don't use the
result_merger to trim since we first need to reorder the rows.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch makes the row limit enforced by the storage_proxy layer.
It adds a row limit to the query_result_merger, useful when merging
results for concurrent queries.
More importantly, it provides guarantees that upper layers may be
relying on implicitly (e.g., the paging code).
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes mutation_query::to_data_query_result() so that it
enforces the row limit alongside the partition limit and the
per-partition limit.
In the following patch, we'll enforce the row limit in an upper layer,
but this lets us optimize the case where only when replica replies.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes a read_command's limits when retrying it, so that
we don't ask for more rows than necessary.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch ensures we keep track of how many partitions we've queried
so we don't ask for more than the number we need.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds a partition count to query::result, filled by the
query::result::builder. The partition count is present whenever the
result carries data, being absent only for the case where the result
contains only a digest.
We also ensure that counts are present for an empty query::result.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes column_family::query() to use the counters in the
builder to determine how many partitions and rows to ask for and also
to implement the stop condition. This saves a continuation to do the
bookkeeping, and allows us to remove data_query_result.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes the query_result_builder to use the counters
provided by the query::result::builder. It also ensures they are kept
current.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes mutation_partition::query_compacted() to count the
number of partitions written to the underlying writer.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds a partition count to the query::result::builder. It is
intended to be incremented by users, and later used to build a
query::result.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds a partition limit to the query_result_merger, useful
when merging results for concurrent queries. This change also makes
the partition limit enforced by the storage_proxy layer, no changes
being needed by the upper layers, namely the Thrift interface.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
We saw the message twice for the same feature check. This is a bit
confusing.
INFO 2016-12-15 11:26:23,993 [shard 0] gossip - Checking if need_features {RANGE_TOMBSTONES} in features {}
INFO 2016-12-15 11:26:23,993 [shard 0] gossip - Checking if need_features {RANGE_TOMBSTONES} in features {}
INFO 2016-12-15 11:26:23,993 [shard 0] gossip - Checking if need_features {LARGE_PARTITIONS} in features {}
INFO 2016-12-15 11:26:23,993 [shard 0] gossip - Checking if need_features {LARGE_PARTITIONS} in features {}
This is because
ss._range_tombstones_feature = gms::feature(RANGE_TOMBSTONES_FEATURE);
ss._large_partitions_feature = gms::feature(LARGE_PARTITIONS_FEATURE);
The first message is printed when gms::feature(RANGE_TOMBSTONES_FEATURE)
is constructed. The second message is printed when the
ss._range_tombstones_feature is copy-constructed.
Add the helper function to enable the a feature and log the feature is
enabled.
When a feature is enabled, we see
INFO 2016-12-15 11:29:32,443 [shard 0] gossip - Feature LARGE_PARTITIONS is enabled
INFO 2016-12-15 11:29:32,443 [shard 0] gossip - Feature RANGE_TOMBSTONES is enabled
in the log.
"This series makes Scylla limit size of query results it produces in case they
grow unreasonably large. This is possible because CQL paging queries do not
guarantee that the returned page is going to have page_size rows and pages
smaller than tha *do not* indicate end of stream. Non-paged queries and Thrift
requests do not have such flexibility and they also get all the requested data
(though their memory usage is still accounted for and may limit paged queries).
There is a maximum result size (1 MB) and all results builders will stop after
reaching it. Moreover, there is a per-shard limitation on the amount of memory
used by all results combined (10%). To avoid tiny results a query has to
reserve (wait if necessary) 4 kB before starting executing, after that it can
consume more memory without any additional waiting provided it is below
individual and shard-local limits.
Enabling the cluster to return less rows than requested also means some changes
for the coordinator. Firstly, if it receives such short result from a replica
retrying it with a larger limit obviously makes no sense whatsoever. Instead,
in such cases the coordinator removes the clustering rows it has incomplate
information about and sends short result back to the client. Moreover, even
if no replica returned short response reconciliation may have made it so. In
this case, the coordinator do not necessairly need to retry the query as well.
Unfortunately, with the current implementation short responses ruin data
queries since they will cause a digest mismatch.
Three new metrics were added:
* database_bytes_total_result_memory -- total memory used by query results
* database_total_operations_short_data_queries -- data queries that were
limited by size, particulary bad as it basically forces coordinator to
retry them as mutation queries
* database_total_operations_short_mutation_queries -- mutation queries limited
by size"
* 'pdziepak/short-paged-reads/v4' of github.com:cloudius-systems/seastar-dev:
storage_proxy: clean up after primary_key introduction
cql3: allow short reads with paged queries
storage_proxy: handle intentional short reads
storage_proxy: make sure coordinator has complete data
storage_proxy: honour partition limit
storage_proxy: use cmd limits to determine that replica reached end
db: add metrics for short reads and memory used for results
data_query: limit result size
mutation_query: limit result size
db: create result_memory_accounters when starting query
query_builder: add partition_slice getter
reconcilable_result: keep result_memory_tracker object
mutation_compactor: honour stop_iteration from consumers
db: add result_memory_limiter
query: add result size limiter
reconcilable_result: properly propagate short_read flag
query_pagers: handle short reads properly
query: allow short reads
serializer_impl: add serializer for bool_class<Tag>
primary_key was introduced as a replacement for
std::pair<dht::decorated_key, std::optional<clustering_key>>. In order
to simplify patch introducing its fields were named 'first' and
'second'. This patch changes the names to something less useless,
removes old row_address alias and removes is_missing_rows() in favour of
primary_key::less_compare_clustering comparator.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
If the result is going to be too large the replica may decide to make it
shorter and coordinator should handle this properly (i.e. do not retry).
Moreover, coordinator could avoid some retries by setting the short_read
flag itself.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
got_incomplete_information() ensures that the coordinator has received
all required data from all replicas.
(see 77dbe3c12f "storage_proxy: fix
reconciliation with limits" for the examples when that may not be the
case).
However, this function is called only if reconciled result has at least
as much rows as the user asked for. This was correct when we had only
total row limit: if the result was shorter than that either all replicas
sent all data they have or the coordinator will retry anyway. However,
since then we got partition limit and per partition row limit and a
request may be limited by one of these while being still below the total
row limit.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
At the moment the coordinator does not care much for the partition
limit. In particular it doesn't check whether after reconciliation the
result still contains enough partitions.
This patch makes it honour the partition limit and increase it in the
retried queries if necessary.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Coordinator may retry a query with larger limits. However, code
determining whether replica has no more data always used the original
limits. This may cause a livelock.
For example, consider cluster having the following partitions (deletions
cover live cells):
node1:
pk=0, v=0
pk=1, v=1
node2
delete pk=0
delete pk=1
pk=2, v=2
pk=3, v=3
Now, if there is a query SELECT * FROM cf LIMIT 2 the first node is
going to send partitions 0 and 1 while second node is going to send 2
and 3 + tombstones for 0 and 1. The coordinator will decide that it
needs to retry the request with larger row limit since node1 may have
some information about partitions 2 and 3 that are newer than what node2
has sent.
However, when the second response arrives node1 will still sent only two
rows since it has no more data. Because the coordinator uses original
row limit it will not notice that this node reached the end and we are
going to get another retry without making any progress.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
This pach ensures than when we start executing a query a minimum result
size is reserved from result_memory_limiter.
Moreover, range queries need a way of merging memory usage information
from different shards.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
This patch introduces an infrastrucutre for limiting result size.
There is a shard-local limit which makes sure that all results combined
do not use more than 10% of the shard memory.
There is also an invidual limit which restricts a result to 4 MB.
In order
In order to avoid sending tiny results there is minimum guaranteed size
(4 kB), which the query needs to reserve before it starts producing the
result.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
reconcilable_result can be merged with another or transformed into
query::result. Make sure that short_read information is never lost.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Currently, the paging implementation assumes that the server retunrs
either as many rows as it was asked for all reached the end. Soon,
that's not going to be true so instead of making any assumptions about
the number of the rows returned use the new "short read" flag to
determine whether there is going to be more data.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
When paging is used the cluster is allowed to return less rows than the
client asked for. However, if such possibility is used we need a way of
telling that to the coordinator and the paging implementation so that
they can differentiate between short reads caused by the replica running
out of data to sent and short reads caused by any other means.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
The test assumed that mutations added to the commitlog are visible to
reads as soon as a new segment is opened. That's not true because
buffers are written back in the background, and new segment may be
active while the previous one is still being written or not yet
synced.
Fix the test so that it expectes that the number of mutations read
this way is <= the number of mutations read, and that after all
segments are synced, the number of mutations read is equal.
Message-Id: <1481630481-19395-1-git-send-email-tgrabiec@scylladb.com>
"The current criteria for memtable flush is not being respected. The
problem is demonstrated to happen when the dirty memory group is over
limit, and so is the system table extra allowance. In that situation,
both the normal region and the system table region will be under
pressure and try to flush.
More specifically, because the normal region inherits from the system
region, if the normal region is under pressure (over the soft limit
threshold), the system region will certainly be as well, even though it
has an extra allowance. This is because after virtual dirty, we start
blocking when we reach half the region, but memory itself can grow up to
100 % of the region. So the total amount of memory used will be
certainly bigger than the system pressure threshold, which is now 50 %
plus the allowance.
To fix that, this patch reworks the flush logic so that the regions are
not dependent on each other.
Fixes#1918"
* 'flush-criteria-v6' of github.com:glommer/scylla:
config: get rid of memtable_total_space
database: rework dirty memory hierarchy
system keyspace: write batchlog mutation in user memory
database: remove flush_token
database: abstract pressure condition notification
database: encapsulate semaphore_units into a flush_permit
database: remove friendship declaration
database: simplify flush_one
database: make memtable_list aware in cases it can't flush
In 7c873f0d (repair: Reduce unnecessary streaming traffic), we optimize
in cases when 1) all the remote nodes has the same checksum and 2) local node
has zero checksum.
In this patch, we make the optimization more generec and cover more cases.
1) With RF = 3, 3 nodes cluster, rm data on node3 then run repair on node2
Before:
INFO 2016-12-09 16:24:31,961 [shard 0] repair - Found differing range (-4091524285777924069, -4086237930244473115]
on nodes {127.0.0.3, 127.0.0.1}, in = {127.0.0.3, 127.0.0.1}, out = {127.0.0.3, 127.0.0.1}
INFO 2016-12-09 16:24:31,963 [shard 0] repair - Found differing range (-609511120964672970, -605253169726090861]
on nodes {127.0.0.1, 127.0.0.3}, in = {127.0.0.1, 127.0.0.3}, out = {127.0.0.1, 127.0.0.3}
INFO 2016-12-09 16:24:31,964 [shard 0] repair - Found differing range (-7655412157560911259, -7652234653747163387]
on nodes {127.0.0.3, 127.0.0.1}, in = {127.0.0.3, 127.0.0.1}, out = {127.0.0.3, 127.0.0.1}
INFO 2016-12-09 16:24:31,965 [shard 0] repair - Found differing range (-4133815130045531703, -4128528774512080749]
on nodes {127.0.0.3, 127.0.0.1}, in = {127.0.0.3, 127.0.0.1}, out = {127.0.0.3, 127.0.0.1}
INFO 2016-12-09 16:24:31,967 [shard 0] repair - Found differing range (-605253169726090861, -600995218487508751]
on nodes {127.0.0.1, 127.0.0.3}, in = {127.0.0.1, 127.0.0.3}, out = {127.0.0.1, 127.0.0.3}
INFO 2016-12-09 16:24:31,968 [shard 0] repair - Found differing range (438510347741343837, 441475345714861354]
on nodes {127.0.0.1, 127.0.0.3}, in = {127.0.0.1, 127.0.0.3}, out = {127.0.0.1, 127.0.0.3}
After:
INFO 2016-12-09 16:30:29,204 [shard 0] repair - Found differing range (-660606535827658284, -656348584589076175]
on nodes {127.0.0.1, 127.0.0.3}, in = {}, out = {127.0.0.3}
INFO 2016-12-09 16:30:29,204 [shard 0] repair - Found differing range (-4234255885181099833, -4228969529647648879]
on nodes {127.0.0.3, 127.0.0.1}, in = {}, out = {127.0.0.3}
INFO 2016-12-09 16:30:29,204 [shard 0] repair - Found differing range (-4228969529647648879, -4223683174114197925]
on nodes {127.0.0.3, 127.0.0.1}, in = {}, out = {127.0.0.3}
INFO 2016-12-09 16:30:29,204 [shard 0] repair - Found differing range (-4223683174114197925, -4218396818580746971]
on nodes {127.0.0.3, 127.0.0.1}, in = {}, out = {127.0.0.3}
INFO 2016-12-09 16:30:29,204 [shard 0] repair - Found differing range (-7728494745277112315, -7725317241463364443]
on nodes {127.0.0.3, 127.0.0.1}, in = {}, out = {127.0.0.3}
INFO 2016-12-09 16:30:29,204 [shard 0] repair - Found differing range (-720217853167807818, -715959901929225709]
on nodes {127.0.0.1, 127.0.0.3}, in = {}, out = {127.0.0.3}
Before, we need to fetch data from both node 1 and node 3 and send data back to node 1 and node 3, i.e., 2 IN, 2 OUT
After, we only need to fetch data from node 3, i.e. 0 IN, 1 OUT
We saved 3X traffic, with higher RF, we can save even more.
2) With RF = 3, 3 nodes cluster, rm data on node3 then run repair on node3
Before:
INFO 2016-12-09 16:20:11,448 [shard 0] repair - Found differing range (-8533861887892628919, -8052600134279395253]
on nodes {127.0.0.1, 127.0.0.2}, in = {127.0.0.1}, out = {}
INFO 2016-12-09 16:20:11,465 [shard 0] repair - Found differing range (7190719703944308372, 7692358524564683543]
on nodes {127.0.0.1, 127.0.0.2}, in = {127.0.0.1}, out = {}
INFO 2016-12-09 16:20:11,486 [shard 0] repair - Found differing range (-3305328316052774469, -2671876682129336880]
on nodes {127.0.0.1, 127.0.0.2}, in = {127.0.0.1}, out = {}
INFO 2016-12-09 16:20:11,494 [shard 0] repair - Found differing range (-2190610927722759275, -1305178847032904465]
on nodes {127.0.0.2, 127.0.0.1}, in = {127.0.0.2}, out = {}
INFO 2016-12-09 16:20:11,518 [shard 0] repair - Found differing range (-4747032371925842389, -4070378863644120252]
on nodes {127.0.0.2, 127.0.0.1}, in = {127.0.0.2}, out = {}
INFO 2016-12-09 16:20:11,519 [shard 0] repair - Found differing range (-1137497074548854552, -592479316010344531]
on nodes {127.0.0.1, 127.0.0.2}, in = {127.0.0.1}, out = {}
After:
INFO 2016-12-09 16:29:22,433 [shard 0] repair - Found differing range (67885601051654285, 447405341661896387]
on nodes {127.0.0.2, 127.0.0.1}, in = {127.0.0.2}, out = {}
INFO 2016-12-09 16:29:22,454 [shard 0] repair - Found differing range (-2190610927722759275, -1305178847032904465]
on nodes {127.0.0.2, 127.0.0.1}, in = {127.0.0.2}, out = {}
INFO 2016-12-09 16:29:22,473 [shard 0] repair - Found differing range (2523396860109747637, 3083778975065200884]
on nodes {127.0.0.2, 127.0.0.1}, in = {127.0.0.2}, out = {}
INFO 2016-12-09 16:29:22,474 [shard 0] repair - Found differing range (-3305328316052774469, -2671876682129336880]
on nodes {127.0.0.1, 127.0.0.2}, in = {127.0.0.1}, out = {}
INFO 2016-12-09 16:29:22,487 [shard 0] repair - Found differing range (-4747032371925842389, -4070378863644120252]
on nodes {127.0.0.2, 127.0.0.1}, in = {127.0.0.2}, out = {}
INFO 2016-12-09 16:29:22,493 [shard 0] repair - Found differing range (-1137497074548854552, -592479316010344531]
on nodes {127.0.0.1, 127.0.0.2}, in = {127.0.0.1}, out = {}
This shows the new more generic methods covers the optimization we had before as well.
Issue #1918 describes a problem, in which we are generating smaller
memtables than we could, and therefore not respecting the flush
criteria.
That happens because group sizes (and limits) for pressure purposes, and
the the soft threshold is currently at 40 %. This causes system group's
soft threshold to be way below regular's virtual dirty limit and close
to regular group's soft threshold. The system group was very likely to
become under soft pressure when regular was because writes to regular
group are not yet throttled when they cross both soft thresholds.
This is a direct consequence of the linear hierarchy between the regions
and to guarantee that it won't happen we would have acqire the semaphore
of all ancestor regions when flushing from a child region. While that
works, it can lead to problems on its own, like priority inversion if
the regions have different priorities - like streaming and regular, and
groups lower in the hierarchy, like user, blocking explicit flushes
from their ancestors
To fix that, this patch reorganizes the dirty memory region groups so
that groups are now completely independent. As a disadvantage, when
streaming happen we will draw some memory from the cache, but we will
live with it for the time being.
Fixes#1918
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Batchlog is a potentially memory-intensive table whose workload is
driven by user needs, not system's. Move it to the user dirty memory
manager.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We had a flush_token structure in addition to the flush_permit because
we needed to keep a pointer to the dirty_memory_manager and apply
changes to the region group upon the region destruction. Since Tomek's
latest series, this is no longer needed and now this structure doesn't
have a place in the world anymore. Simplify the code by removing it.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Done in a separate patch to reduce clutter in the main patch.
Soon we'll be testing for one more condition.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We will soon need to hold more than a semaphore_units<> object per
flush, potentially.
Preparation patch for that.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
flush_one has to make sure that we're using the correct
dirty_memory_manager object, because we could be flushing from a region
group different than the one the flush request originated.
It's simpler to just assume flush_one will be dealing with the right
object, and use a different object instead of "this" when calling it.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Some of our CFs can't be flushed. Those are the ones who are not marked
as having durable writes. We treat them just the same from the point of
view of the flush logic, but they provide a function that doesn't do
anything and just returns right away.
We already had troubles with that in the past, and that also poses a
problem for an upcoming patch reworking the flush memtable pick
criteria.
It's easier, simpler, and cleaner, to just make the memtable_list aware
it can't flush. Achieving that is also not very complicated: we just
need a special constructor that doesn't take a seal function and then we
make sure that it is initialized to an empty std::function
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Currently, we weren't completing a query as early as possible if it
reached the partition limit, we instead had to wait until reaching the
end of the specified partition ranges. This patches fixes that by
including a check to the partition limit in the termination condition.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20161213114559.26438-1-duarte@scylladb.com>
column_mapping is not safe to access across shards, because data_type
is not safe to access. One of the manifestation of this is that
abstract_type::is_value_compatible_with() always fails if the two
types belong to different shards.
During replay, column_mapping lives on the replaying shard, and is
used by converting_mutation_partition_applier against the schema on
the target shard. Since types in the mapping will be considered
incompatible with types in the schema, all cells will be dropped.
Fix by using column_mapping in a safe way, by copying it to the target
shard if necessary. Each shard maintains its own cache of column
mappings.
Fixes#1924.
Message-Id: <1481310463-13868-1-git-send-email-tgrabiec@scylladb.com>
* seastar 0773e98...6fbd792 (2):
> tls: Only run our "verify" function in client session
> Merge "Clean the metric definition" from Amnon
Includes patch from Amnon adjusting the metrics registration due to seastar
API changes.
* seastar 0a74317...0773e98 (6):
> tls: Add support for client cetrificate verification & priority strings
> semaphore: add consume_units
> semaphore: add available_units()
> thread: check need_preempt for threads in a scheduling group as well
> tutorial: fix semaphore example, and text
> stop_iteration: add && and || operators
"This series:
- We can make reader with ranges
- Fix possible use after free of 'si'
- Streaming ranges now are sorted and merged
- Fix shard_begin shard_end end loop in both streaming and repair"
A range now alternates between different shards: the first part of the
range goes to shard X, the next to shard X+1, but after a while we go
back to shard X. So we can't do a simple loop between shard_begin and
shard_end.
Fix by using the newly introduced dht::split_range_to_shards
Use the cf.make_streaming_reader with ranges to simplify the code a bit.
Now that we have the new interface to make readers with ranges, we can
simplify the code a lot.
1) Less readers are needed
before: number of ranges of readers
after: smp::count readers at most
2) No foreign_ptr is needed
There is no need to forward to a shard to make the foreign_ptr for
send_info in the first phase and forward to that shard to execute the
send_info in the second phase.
3) No do_with is needed in send_mutations since si now is a
lw_shared_ptr
4) Fix possible user after free of 'si' in do_send_mutations
We need to take a reference of 'si' when sending the mutation with
send_stream_mutation rpc call, otherwise:
msg1 got exception
si->mutations_done.broken()
si is freed
msg2 got exception
si is used again
The issue is introduced in dc50ce0ce5 (streaming: Make the mutation
readers when streaming starts) which is master only, branch 1.5 is not
affected.
Allow to make a streaming reader with a vector of ranges in addition to
a single range. This will be used soon in following streaming patch.
We can make the reader more efficient later.
This patch fixes a typo in i_partitioner::tri_compare() where we were
using std::max instead of std::min, thus avoiding accessing random
memory and getting random results.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20161211165043.17816-1-duarte@scylladb.com>
This patch adds uuid file support for ubuntu system. It also split the
behaviour between restart and daily checks. The first run in r mode and
the second in d mode.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Allows scylla-housekeeping getting the uuid from a file instead of the
command line.
If the file is missing no uuid will be used.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
"Function to calculate maximum purgeable timestamp is made 10 times faster when
compacting sstables overlap with 10% of all sstables.
That's possible with an incremental selector that will incrementally select
sstables based on key being compacted.
Currently, we iterate through all non-compacting sstables and consult their
bloom filter to determine max purgeable timestamp, and that will be very
expensive for compactions that are frequently deciding whether or not to purge
tombstones."
* 'filter_overhead_fix_v4' of github.com:raphaelsc/scylla:
compaction: reduce bloom filter overhead with incremental selector
tests: add test for sstable set's incremental selector
sstable_set: introduce incremental selector
compatible_ring_position: add function to return token
The procedure to calculate max purgeable timestamp is optimized
by only visiting sstables that overlap with key being currently
compacted. That's done using incremental sstable selector.
Function to calculate maximum purgeable timestamp is made 10 times
faster when compacting sstables overlap with 10% of all sstables.
Fixes#1322.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Incrementally select sstables from sstable set using token
in ascending order.
For leveled strategy, it returns all sstables that belong
to current interval. For other strategies, it just return
all sstables from the set.
Useful for compaction which needs all sstables that overlap
with key being currently compacted to calculate maximum
purgeable timestamp.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The semaphore future may be unavailable for many reasons. Specifically,
if the task quota is depleted right between sem.wait() and the .then()
clause in get_units() the resulting future won't be available.
That is particularly visible if we decrease the task quota, since those
events will be more frequent: we can in those cases clearly see this
counter going up, even though there aren't more requests pending than
usual.
This patch improves the situation by replacing that check. We now verify
whether or not there are waiters in the semaphore.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <113c0d6b43cd6653ce972541baf6920e5765546b.1481222621.git.glauber@scylladb.com>
When row cache is disabled, update_cache() will do nothing to the
memtable. Active readers may keep the memtable alive for unbounded
amount of time, preventing it from going away. This doesn't play well
with virtual dirty accounting. Soon before calling update_cache(), the
memory which was subtracted during flush is added back to the amount
of virtual dirty memory. If there was write pressure all along, we
will be at the dirty memory limit. When we give back subtracted memory
this will put virtual dirty way above the limit. This will stall all
writes until another memtable flush drags virtual dirty down or
readers finally release the memtable. We want to prevent upward
jumps of virtual dirty.
First part of the fix is to ensure that as long as the memtable's
region is in the dirty group, we will not revert flushed memory. This
must happen synchronously from region's memory being removed from the
group in order to prevent upward virtual dirty jumps. To make this
easier, tracking of flushed memory was moved to the memtable object.
Another part of the fix is to gradually clear the memtable when cache
is disabled in a similar fashion as when it's moved to cache. This
ensures that the actual memory held by memtable's region is released
sooner than it dies.
Refs #1879
"Due to my misreading of Cassandra code, I thought it would ignore new
components in the Statistics component; however, it doesn't, and the change
(introduced in bdd11648ac ("sstables: add
intra-node sharding metadata") breaks sstable2json and likely any
Cassandra code that touches sstables.
To fix, move the sharding data into a new component ("Scylla.db"), which
Cassandra does ignore. The new component is designed to be extensible so
we don't experience the same issue later on."
* tag 'asias/repair/subranges/refactor_fix/v1' of github.com:cloudius-systems/seastar-dev:
repair: Limit the number of sub ranges
repair: Use estimated_keys_for_range in repair_cf_range
repair: Extract the target_partitions into repair_info class
repair: Put request_transfer_ranges into repair_info class
repair: Introduce check_failed_ranges helper
repair: Introduce do_streaming helper
repair: Make the neighbors const reference
repair: Introduce repair_info
repair: Attach the repair id in the stream plan name
The problem is that replay will unlink any segments which were on disk
at the time the replay starts. However, some of those segments may
have been created by current node since the boot. If a segment is part
of reserve for example, it will be unlinked by replay, but we will
still use that segment to log mutations. Those mutations will not be
visible to replay after a crash though.
The fix is to record preexisting segents before any new segments will
have a chance to be created and use that as the replay list.
Introduced in abe7358767.
dtest failure:
commitlog_test.py:TestCommitLog.test_commitlog_replay_on_startup
Message-Id: <1481117436-6243-1-git-send-email-tgrabiec@scylladb.com>
- Add "coordinator" and "replica" categories
- Use a new seastar/metrics_registration framework
* 'rearrange-storage-proxy-stats-v4' of github.com:cloudius-systems/seastar-dev:
service::storage_proxy: rework the collectd counters registration
service/storage_proxy: regroup collectd statistics
The Cassandra derived sstable tools (and likely Cassandra itself) object to
a new sub-component in the Statistics component; create a new Scylla
component instead to host this data.
Allow declaring discriminated unions (with an enum type as the
discriminant and any sstable serializable type as a value) and sets
of these unions, with the disciminant as the key. Parsers and writers
are auto-generated.
Currently housekeeping timer won't be reset when we restart scylla-server.
We expect the service to be run at each start, it will be consistent with
upstart script in Ubuntu 14.04
When we restart scylla-server, housekeepting timer will also be restarted,
so let's replace "OnBootSec" with "OnActiveSec".
Fixes: #1601
Signed-off-by: Amos Kong <amos@scylladb.com>
Message-Id: <a22943cc11a3de23db266c52fd476c08014098c4.1480607401.git.amos@scylladb.com>
To reduce duplicated code and simplified scripts introduce scylla_lib.sh
for shellscripts which provides functions to classify distributions,
and load all sysconfig files.
This also fixes script bugs to misdetect Debian and RHEL.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1480667672-9453-2-git-send-email-syuu@scylladb.com>
Since Ubuntu 15.10/16.04 still uses Upstart to manage GUI session (not as init), when we directly launch Scylla on Ubuntu's GUI Terminal(not using systemctl or initctl), raise(SIGSTOP) mistakenly calls (Because GUI session has "UPSTART_JOB" environment variable, won't happen when running Scylla as systemd service).
To avoid this, we need to verify UPSTART_JOB == "scylla-server".
If it's part of GUI session UPSTART_JOB has to be "unity7", we need to avoid raise(SIGSTOP) in that case.
Fixes#1199
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1480620421-28967-1-git-send-email-syuu@scylladb.com>
So that memory is released gradually (impacting latency less) and
sooner than when memtable is destroyed. Active readers may keep the
memtable alive for unbounded amount of time.
Refs #1879
When memtable is flushing, it subtracts _flushed_memory from groups's
size to gradually allow more writes. Ideally _flushed_memory would be
equal to region's size when flush ends, so the group's size would
reach zero. When the memtable and its region are gone the group size
should remain the same as after the flush. This is ensured by adding
back _flushed_memory to group's size right before the region is
removed from the group.
Calling clear() before region is removed from the group breaks the
accounting because it will shrink the region, but will not affect the
amount of memory subtracted due to _flushed_memory. So group's size
would decrease more than we want (twice the region's size). The fix is
to change clear() so that it reverts _flushed_memory by the amount by
which the region size is reduced. This will keep the groups's size
constant as long as _flushed_memory > 0.
The implementation assumes that memtable's region group is owned by
dirty_memory_manager, and tries to obtain a reference to it like this:
boost::intrusive::get_parent_from_member(_region.group(), &dirty_memory_manager::_region_group));
This is undefined behavior when the region's group does not come from
dirty manager. It's safer to be explicit about this dependency by
taking a reference to dirty_memory_manager in the constructor.
If exception is triggered early in boot when doing an I/O operation,
scylla will fail because io checker calls storage service to stop
transport services, and not all of them were initialized yet.
Scylla was failing as follow:
scylla: ./seastar/core/sharded.hh:439: Service& seastar::sharded<Service>::local()
[with Service = gms::gossiper]: Assertion `local_is_initialized()' failed.
Aborting on shard 0.
Backtrace:
0x000000000048a2ca
0x000000000048a3d3
0x00007fc279e739ff
0x00007fc279ad6a27
0x00007fc279ad8629
0x00007fc279acf226
0x00007fc279acf2d1
0x0000000000c145f8
0x000000000110d1bc
0x000000000041bacd
0x00000000005520f1
0x00007fc279aeaf1f
Aborted (core dumped)
Refs #883.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Signed-off-by: Asias He <asias@scylladb.com>
Message-Id: <963f7b0f5a7a8a1405728b414a7d7a6dccd70581.1479172124.git.asias@scylladb.com>
A range is diveded into N sub ranges so that each sub range contains 100
partitions. So N depends on the number of partitions in that range. N
can grow unbounded and the memory usage of vector to hold these sub
ranges can go unbouded.
Limit the max number of sub ranges a range can divided into.
The downside is that the limited sub range will make we include more
partitions in the checksum.
Fixes#1917
Use the new seastar's metrics_registration framework:
- Change the registration syntax.
- Add a long description for each counter.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Instead of putting all statistics under the same "storage_proxy" category
separate them into 2 groups according to where the corresponding counters
are updated:
- "storage_proxy_replica"
- "storage_proxy_coordinator"
Fixes#1763
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
When requests hit the commitlog, each of them will be assigned a replay
position, which we expect to be ordered. If reorders happen, the request
will be discarded and re-applied. Although this is supposed to be rare,
it does increase our latencies, specially when big requests are
involved. Processing big requests is expensive and if we have to do it
twice that adds to the cost.
The commitlog is supposed to issue replay positions in order, and it
coudl be that the code that adds them to the memtables will reorder
them. However, there is one instance in which the commitlog will not
keep its side of the bargain.
That happens when the reserve is exhausted, and we are allocating a
segment directly at the same time the reserve is being replenished. The
following sequence of events with its deferring points will ilustrate
it:
on_timer:
return this->allocate_segment(false). // defer here // then([this](sseg_ptr s) {
At this point, the segment id is already allocated.
new_segment():
if (_reserve_segments.empty()) {
[ ... ]
return allocate_segment(true).then ...
At this point, we have a new segment that has an id that is higher than
the previous id allocated.
Then we resume the execution from the deferring point in on_timer():
i = _reserve_segments.emplace(i, std::move(s));
The next time we need to allocate a segment, we'll pick it from the
reserve. But the segment in the reserve has an id that is lower than the
id that we have already used.
Reorders are bad, but this one is particularly bad: because the reorder
happens with the segment id side of the replay position, that means that
every request that falls into that segment will have to be reinserted.
This bug can be a bit tricky to reproduce. To make it more common, we
can artificially add a sleep() fiber after the allocate_segment(false)
in on_timer(). If we do that, we'll see a sea of reinsertions going on
in the logs (if dblog is set to debug).
Applying this patch (keeping the sleep) will make them all disappear.
We do this by rewriting the reserve logic, so that the segments always
come from the reserve. If we draw from a single pool all the time, there
is no chance of reordering happening. To make that more amenable, we'll
have the reserve filler always running in the background and take it out
of the timer code.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <49eb7edfcafaef7f1fdceb270639a9a8b50cfce7.1480531446.git.glauber@scylladb.com>
The information (last compacted keys) is lost after node is restarted
or schema is updated, which causes strategy to be rebuilt.
We need it for strategy to guarantee uniform distribution of token
range across sstables, or we could end up with 1 sstable of level L
overlapping with lots of sstables of level L+1, and that results in
a compaction of undesired length.
That information can be generated from scratch by getting last key
of newest sstable in each level > 0.
Fixes#1906.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <35ebd15977d5a8418239febb160c796cdc0e98fa.1480533805.git.raphaelsc@scylladb.com>
Streaming memtable have a delayed mode where many flushes are coalesced
together into one, with the actual flush happening later and propagated
to all the previous waiters.
However, the timer that triggers the actual flush was not using the
newly introduced flush infrastructure. This was a minor problem because
those flushes wouldn't try to take the semaphore, and so we could have
many flushes going on at the same time.
What was a potential performance issue became a correctness issue when
we moved the reversal of the dirty memory accounting out of
revert_potentially_cleaned_up_memory() into remove_from_flush_manager().
Since the latter is only called through the flush infrastructure, it
simply wasn't called. So the deferral of the reversal exposed this bug.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <0d5755375bc27524b8cfb9970c76d492b14d9eea.1480522742.git.glauber@scylladb.com>
This fix splits build_ami.sh --repo to three different options:
--repo-for-install is for Scylla package installation, only valid
during AMI construction.
--repo-for-update will be stored at /etc/yum.repos.d/scylla.repo, to
receive update package on AMI.
--repo is both, for installation and update.
Fixes#1872
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1480438858-6007-1-git-send-email-syuu@scylladb.com>
"The goal of this series is to prevent unbounded memory use
in cases when requests are timing out. Write requests which timed
out may still occupy memory for a while because of local mutation
application. This memory is not accounted for and can build up.
First part of the fix changes local mutation application so that it times out
at about the same time as the request handler. Then the life
time of the request handler is extended to cover any background activity
of that request which hasn't timed out yet. This has two main effects:
(1) by timing out local writes we prevent build up of background activity
for timed out requests
(2) we ensure that memory used by background activity is not left
behind unaccounted for. This will prevent CQL server from admitting
more requests than memory usage limit allows.
Fixes #1756."
* tag 'tgrabiec/prevent-oom-on-timeouts-v5' of github.com:cloudius-systems/seastar-dev:
storage_proxy: Do not flood logs with timeout errors
database: Add counter for timed out writes
storage_proxy: Delay timeout response until background work ceases
storage_proxy: Propagate timeout to local writes
storage_proxy: Use shared ownership for abstract_write_response_handler
storage_proxy: Add counter for all alive write handlers
db: Allow writes to be timed out
db: Introduce counters for failed reads and writes
commitlog: Allow allocations to be timed out
utils/logalloc: Add ability to timeout run_when_memory_available() task
utils/flush_queue: Add ability to wait with a timeout
Timeout errors are flooding the log after local mutate can time
out. We don't log remote mutate timeouts, so for consistency we won't
log local ones as well.
There is a database counter for timed out writes which can be
consulted in order to check if they're occuring.
Perhaps this would be better solved by a generic log message
throttling/coalescing mechanism, but that's not ready yet.
Write requests which timed out may still occupy memory for a while due
to local write. It should time out soon as well but there is a time
window in which it has not yet. If we don't delay timeout response,
the request would be seen as not consuming any memory too early. This
in turn would cause the CQL server to allow more requests than we
want. In some cases causing OOM or exceeding memory limits and causing
excessive cache eviciton.
Fixes#1756.
Currently the counter uses _response_handlers.size(), but after later
patches we may have an active (timed out) write with no response
handler, so count live instances instead.
Problem will cause size tiered to return small jobs when there are
more than max_threshold sstables of similar size. For example, if
max_threshold is 32, and there are 36 sstables of similar size,
strategy will only return 4 sstables to be compacted. That's because
we incorrectly create a new bucket when it meets the max threshold.
What we should do is to allow buckets to grow beyond max threshold
and trim them when selecting the most suitable one for compaction.
Important to mention that estimation for size tiered will now
work better when there are more than max_threshold sstables of
similar size.
Fixes#1901.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <080bad70d6cb86eaf52ac1bdd6765ac47aab5b03.1478316140.git.raphaelsc@scylladb.com>
This reverts commit 0e9b75d406.
commitlog_test fails with this:
Running 14 test cases...
ERROR 2016-11-28 20:48:00,565 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:00,578 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:10,591 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:20,601 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
tests/commitlog_test.cc(203): fatal error in "test_commitlog_discard_completed_segments": critical check dn <= nn failed
ERROR 2016-11-28 20:48:20,645 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:20,837 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
WARN 2016-11-28 20:48:20,838 [shard 0] commitlog - Exception in segment reservation: std::system_error (error system:2, No such file or directory)
ERROR 2016-11-28 20:48:20,952 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:31,064 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:31,083 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:31,098 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:31,111 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
ERROR 2016-11-28 20:48:31,113 [shard 0] commitlog - Segment reserve is full! Ignoring and trying to continue, but shouldn't happen
WARN 2016-11-28 20:48:31,116 [shard 0] commitlog - Could not allocate 16388 k bytes output buffer (16388 k required)
*** 1 failure detected in test suite "tests/commitlog_test.cc"
WARN 2016-11-28 20:48:31,117 [shard 0] commitlog - Exception in segment reservation: std::system_error (error system:2, No such file or directory)
New sstables are loaded and added in parallel, meaning that scylla can
potentially return stale data if a new sstable containing a tombstone
wasn't loaded yet. Compaction should also not run until all new sstables
are added for similar reasons.
Fix is about separating blocking and non-blocking steps to allow
atomic add of multiple new sstables.
Fixes#1368.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <14283b8a4a69127071d1fabef320a93c91817ec2.1480356073.git.raphaelsc@scylladb.com>
When requests hit the commitlog, each of them will be assigned a replay
position, which we expect to be ordered. If reorders happen, the request
will be discarded and re-applied. Although this is supposed to be rare,
it does increase our latencies, specially when big requests are
involved. Processing big requests is expensive and if we have to do it
twice that adds to the cost.
The commitlog is supposed to issue replay positions in order, and it
coudl be that the code that adds them to the memtables will reorder
them. However, there is one instance in which the commitlog will not
keep its side of the bargain.
That happens when the reserve is exhausted, and we are allocating a
segment directly at the same time the reserve is being replenished. The
following sequence of events with its deferring points will ilustrate
it:
on_timer:
return this->allocate_segment(false). // defer here // then([this](sseg_ptr s) {
At this point, the segment id is already allocated.
new_segment():
if (_reserve_segments.empty()) {
[ ... ]
return allocate_segment(true).then ...
At this point, we have a new segment that has an id that is higher than
the previous id allocated.
Then we resume the execution from the deferring point in on_timer():
i = _reserve_segments.emplace(i, std::move(s));
The next time we need to allocate a segment, we'll pick it from the
reserve. But the segment in the reserve has an id that is lower than the
id that we have already used.
Reorders are bad, but this one is particularly bad: because the reorder
happens with the segment id side of the replay position, that means that
every request that falls into that segment will have to be reinserted.
This bug can be a bit tricky to reproduce. To make it more common, we
can artificially add a sleep() fiber after the allocate_segment(false)
in on_timer(). If we do that, we'll see a sea of reinsertions going on
in the logs (if dblog is set to debug).
Applying this patch (keeping the sleep) will make them all disappear.
We do this by rewriting the reserve logic, so that the segments always
come from the reserve. If we draw from a single pool all the time, there
is no chance of reordering happening. To make that more amenable, we'll
have the reserve filler always running in the background and take it out
of the timer code.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <2606b97df39997bcf3af84a23adf17e094ffb0b8.1480107174.git.glauber@scylladb.com>
"We currently write the size_estimates system table for every schema
on a periodic basis, currently set to 5 minutes, which can interfere
with an ongoing workload.
This patchset virtualizes it such that queries are intercepted and we
calculate the results on the fly, only for the ranges the caller is interested in.
Fixes#1616"
* 'virtual-estimates/v4' of github.com:duarten/scylla:
size_estimates_virtual_reader: Add unit test
db: Delete size_estimates_recorder
size_estimates: Add virtual reader
column_family: Add support for virtual readers
storage_service: get_local_tokens() returns a future
nonwrapping_range: Add slice() function
range: Find a sequence's lower and upper bounds
system_keyspace: Build mutations for size estimates
size_estimates: Store the token range as bytes
range_estimates: Add schema
murmur3_partitioner: Convert maximum_token to sstring
When we finish writing a memtable, we revert the dirty memory charges
immediately. When we do that, dirty memory will grow back to what it
was, and soon (we hope) will go down again when we release the requests
for real.
During that time, we may not accept new requests. Sealing can take a
long time, specially in the face of Linux issues like the ones we have
seen in the past. It also will take proportionally more time if the
SSTables end up being small, which is a possibility in some scenarios.
This patch changes the dirty_memory_manager so that the charges won't be
reverted right after we finish the flush. Rather, we will hold on to it,
and revert it right before we update the cache. We don't need to do it
for all classes of memtable writes, because after we finish flushing,
flush_one() will destroy the hashed element anyway.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <2d5a8f6ca57d5036f4850ac163557bca59b8063d.1480004384.git.glauber@scylladb.com>
GCC 5.3.1 was unable to convert bound to optional<bound>.
sstables/sstables.cc:2494:123: error: no matching function for call to
‘nonwrapping_range<dht::ring_position>::nonwrapping_range(dht::ring_position,
dht::ring_position)’
(dtr.right.exclusive ? dht::ring_position::starting_at :
dht::ring_position::ending_at)(std::move(t2)));
In file included from ./dht/i_partitioner.hh:52:0,
from ./query-request.hh:28,
from ./clustering_key_filter.hh:27,
from sstables/sstables.hh:35,
from sstables/sstables.cc:38:
./range.hh:441:14: note: candidate: nonwrapping_range<T>::nonwrapping_range(
const wrapping_range<U>&) [with T = dht::ring_position]
explicit nonwrapping_range(const wrapping_range<T>& r)
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <95bbf984cd73a61739c8da99cf6cd5e94f1d1457.1479954360.git.raphaelsc@scylladb.com>
Since not all distributions have a version of LZ4 with
LZ4_compress_default(), we use it conditionally.
This is specially important beginning with version 1.7.3 of LZ4,
which deprecates the LZ4_compress() function in favour of
LZ4_compress_default() and thus prevents Scylla from compiling
due to the deprecated warning.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20161124092339.23017-1-duarte@scylladb.com>
In Thrift, SliceRange defines a count that limits the number of cells
to return from that row (in CQL3 terms, it limits the number of rows
in that partition). While this limit is honored in the engine, the
Thrift layer also applies the same limit, which, while redundant in
most cases, is used to support the get_paged_slice verb.
Currently, the limit is not being reset per Thrift row (CQL3
partition), so in practice, instead of limiting the cells in a row,
we're limiting the rows we return as well. This patch fixes that by
ensuring the limit applies only within a row/partition.
Fixes#1882
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <20161123220001.15496-1-duarte@scylladb.com>
With the default value of 12, a node's range is partitioned into
4096 * smp::count sub-ranges which are queried sequentually for a range
scan. If the number of rows in the table is smaller than the required
result size, we will query all of them. This can take so long that we
time out.
A better fix is to query multiple sub-ranges in parallel and merge them,
but for that we need to resurrect the non-sequential merger.
"Clusters with a large number of nodes, or a low number of vnodes, and a
high number of shards, or a combination, suffer from an aliasing problem:
both vnodes and intra-node sharding consider the most significant bits
to select the owning node and owning shard respectively. Since the same
bits are used for both, a low number of vnodes leads to some shards
being overcommitted relative to others.
This series fixes the problem by sharding on bits 0:47 of the token
(murmur3 partitioner only), leaving the most significant 12 bits for
vnodes. Simulation shows that this value provides reasonable sharding
for 100-node, 30-shard clusters.
In order to prevent re-sharding sstables on each boot, token ranges for
the range are stored in a new sub-component of the sstable Statistics
component. With the default 12 ignored bits we have 4096 token ranges
for non-Level-compacted SSTables, which takes some space but is still
reasonable.
Fixes #1277."
Sharding on the most significant token bits aliases with the vnode mechanism,
which also uses the most significant bits; this requires a huge number of
vnodes to achieve good sharding.
This patch teaches the murmur3 partitioner to ignore the most significant
N bits when calculating a token's hard, so we use token bits which still have
some entropy. In effect, with changes the token range layout from
shard 0
shard 1
...
shard S-1
to
shard 0
shard 1
...
shard S-1
shard 0
shard 1
...
shard S-1
...
shard 0
shard 1
...
shard S-1
Where the number of repetitions of the block is 2^(ignored msb bits).
For compatibility, the default is zero ignored bits, matching the pre-patch
state, until we wire things up.
Instead of calculating the owning shard from the sstable's partition
key range, delegate to the new sstable method for getting owning shard
infomation. This insulates us from changes in the sharding algorithm.
When we load an sstable, we don't know beforehand which shards it belongs
to; we don't want to open it until we do. Add a method that allows us
to read just the sharding data, without opening anything else.
Add a metadata component that describes token ranges that are spanned by
this sstable. With the current sharding algorithm, where each shard owns
a single token range, the first/last partition key is sufficient to
describing sharding information, but for multi-range algorithms, this
is not sufficient.
We have a semaphore controlling the amount of background work generated
by the memtable flush process. However, because we are not moving it
inside the memtable post-flush continuation, the units are being
released when we star the flush and not when we finish it.
That's not the intended behavior and that can cause flushes to
accumulate.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <b7dc1866ed3473b9b1862c433d59c5ebd8575dbc.1479839600.git.glauber@scylladb.com>
Instead of calculating the offset for each statistic component manually,
use a loop to iterate over all components, accumulating the offset as we
go along.
Introduce a new function that reuses the file_writer code to compute
the serialized size of an sstable object, by serializing it into memory
and discarding the result.
write() doesn't need to change its input; so change it to const.
The only snag is that describe_type() isn't and can't be made const-correct,
so cheat when it is called and const_cast the input.
This helps in writing a generic serialized_size() that is const correct,
in the next patch.
Recently we have changed our shutdown strategy to wait for the
_request_controller semaphore to make sure no other allocations are
in-flight. That was done to fix an actual issue.
The problem is that this wasn't done early enough. We acquire the
semaphore after we have already marked ourselves as _shutdown and
released the timer.
That means that if there is an allocation in flight that needs to use a
new segment, it will never finish - and we'll therefore neve acquire
the semaphore.
Fix it by acquiring it first. At this point the allocations will all be
done and gone, and then we can shutdown everything else.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <5c2a2f20e3832b6ea37d6541897519a9307294ed.1479765782.git.glauber@scylladb.com>
storage_proxy has an optimization where it tries to query multiple token
ranges concurrently to satisfy very large requests (an optimization which is
likely meaningless when paging is enabled, as it always should be). However,
the rows-per-range code severely underestimates the number of rows per range,
resulting in a large number of "read-ahead" internal queries being performed,
the results of most of which are discarded.
Fix by disabling this code. We should likely remove it completely, but let's
start with a band-aid that can be backported.
Fixes#1863.
Message-Id: <20161120165741.2488-1-avi@scylladb.com>
We current pass a region group to the memtable, but after so many recent
changes, that is a bit too low level. This patch changes that so we pass
a memtable list instead.
Doing that also has a couple of advantages. Mainly, during flush we must
get to a memtable to a memtable_list. Currently we do that by going to
the memtable to a column family through the schema, and from there to
the memtable_list.
That, however, involves calling virtual functions in a derived class,
because a single column family could have both streaming and normal
memtables. If we pass a memtable_list to the memtable, we can keep
pointer, and when needed get the memtable_list directly.
Not only that gets rid of the inheritance for aesthetic reasons, but
that inheritance is not even correct anymore. Since the introduction of
the big streaming memtables, we now have a plethora of lists per column
family and this transversal is totally wrong. We haven't noticed before
because we were flushing the memtables based on their individual sizes,
but it has been wrong all along for edge cases in which we would have to
resort to size-based flush. This could be the case, for instance, with
various plan_ids in flight at the same time.
At this point, there is no more reason to keep the derived classes for
the dirty_memory_manager. I'm only keeping them around to reduce
clutter, although they are useful for the specialized constructors and
to communicate to the reader exactly what they are. But those can be
removed in a follow up patch if we want.
The old memtable constructor signature is kept around for the benefit of
two tests in memtable_tests which have their own flush logic. In the
future we could do something like we do for the SSTable tests, and have
a proxy class that is friends with the memtable class. That too, is left
for the future.
Fixes#1870
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <811ec9e8e123dc5fc26eadbda82b0bae906657a9.1479743266.git.glauber@scylladb.com>
Now that access to the size_estimates system is virtualized, we no
longer need the recorder.
Fixes#1616
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch add a virtual mutation_reader so that queries
to the size_estimates system table are handled by the engine
without needing to perform any IO.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Virtual readers allow queries to selected tables, usually system
tables, to be answered by the engine. This is useful for tables which
aren't written by users and whose contents can be calculated on
demand.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes the get_local_tokens() function in storage_service
to return a future instead of requiring running under a seastar::thread.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch add the slice() function to nonwrapping range, which uses
its bounds to slice an input sequence.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch extracts a pair of functions from mutation_partition to
calculate the lower and upper bounds of a sequence from a
nonwrapping_range.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds a function to system_keyspace responsible for creating
a mutation to a partition of the size_estimates system table from a
set of range_estimates.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch changes the range_estimates struct so that the tokens are
represented as utf8 encoded bytes. This will make future patches
require less conversions.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch ensures we can convert the maximum_token to an sstring.
For Cassandra, the minimum and maximum tokens have the same
representation. So, we use the string representation of the
maximum_token for the maximum_token.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
"This series is just a cleanup which intention is to deal with all
confusion related to the way T::memory_usage() functions work.
* T::memory_usage() which returned external memory usage are renamed
to T::external_memory_usage()
* T::memory_usage() is introduced where needed to avoid repeating
sizeof(T) + T::external_memory_usage()"
Paweł Dziepak (6):
rename memory_usage() to external_memory_usage() where applicable
streamed_mutation: add memory_usage() to mutation fragment types
keys: add memory_usage()
partition_snapshot_accounter: use range_tombstone::memory_usage()
mutation_rebuilder: use memory_usage()
frozen_mutation: use memory_usage()
In the last iterations of this patchset, we have moved explicit flushes
to acquire the semaphore directly and the coalescing inside the
memtable_list. As a result, we are no longer keeping any kind of action
for them inside the condition variable. Checking for them has no longer
a purpose.
This is a cleanup patch that remove does checks.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <732676ccfe4ac93eb57aa799ec94b841499a01a6.1479500646.git.glauber@scylladb.com>
If a Column Family is non-durable, then its flushes will never create a
memtable flush reader. Our current flush logic depends on that being
created and destroyed to release the semaphore permits on the flush.
We will remove the permits ourselves it there is an exception, but not
under normal circumnstances. Given this issue, however, it would be more
adequate to always try to remove the permits after we flush. If the
permits were already removed by the flush reader, then this test will
just see that the permit is not in the map and return. But if it is
still there, then it is removed.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <049334c3b4bef620af2c7c045e6c84347dcf9013.1479498026.git.glauber@scylladb.com>
This patch introduces memory_usage() to static_row, clustering_row and
range_tombstone so that we can avoid repeating sizeof(T) +
x.external_memory_usage().
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Renaming the function to external_memory_usage() makes it clear that
sizeof(T) is not included, something that was a source of confusion in
the past.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
"Fixes #1856
Commitlog replay reads are being issued without a priority. That means
they will lose to compaction every time."
* 'issue-1856-v2' of github.com:glommer/scylla:
commitlog: use read ahead for replay requests
commitlog: use commitlog priority for replay
commitlog: close replay file
We supported install CentOS7 .rpm on RHEL7, but we haven't supported
building on RHEL7, since there is little difference between CentOS,
and that causes build error.
This patch fixes the error, now we can produce .rpm for RHEL7 wihout
using CentOS.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1479431134-8032-1-git-send-email-syuu@scylladb.com>
This patch addresses post-merge follow up comments by Tomek.
Basically, what we do is:
- we don't need to signal() from remove_from_flush_manager(), because
the explicit flushes no longer wait on the condition variable. So we
don't.
- We now wait on the stop() flushes (regardless of their return status)
so we can make sure that the _flush_queue will indeed be done with.
- we acquire the semaphore before shutting down the dirty_memory_manager
to make sure that there are no pending flushes
- the flush manager that holds the semaphore has to match in the exception
handler
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <a23ab5098934546c660a08de64cd9294bb3a2008.1479400239.git.glauber@scylladb.com>
Aside from putting the requests in the commitlog class, read ahead
will help us going through the file faster.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Right now replay is being issued with the standard seastar priority.
The rationale for that at the time is that it is an early event that
doesn't really share the disk with anybody.
That is largely untrue now that we start compactions on boot.
Compactions may fight for bandwidth with the commitlog, and with such
low priority the commitlog is guaranteed to lose.
Fixes#1856
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Replay file is opened, so it should be closed. We're not seeing any
problems arising from this, but they may happen. Enabling read ahead in
this stream makes them happen immediately. Fix it.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
The leakage results in deleted sstables being opened until shutdown, and disk
space isn't released. That's because column_family::rebuild_sstable_list()
will not remove reference to deleted sstables if an exception was triggered in
sstables::delete_atomically(). A sstable only has its files closed when its
object is destructed.
The exception happens when a major compaction is issued in parallel to a
regular one, and one of them will be unable to delete a sstable already deleted
by the other. That results in remove_by_toc_name() triggering boost::filesystem
::filesystem_error because TOC and temporary TOC don't exist.
We wouldn't have seen this problem if major compaction were going through
compaction manager, but remove_by_toc_name() and rebuild_sstable_list() should
be made resilient.
Fixes#1840.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <d43b2e78f9658e2c3c5bbb7f813756f18874bf92.1479390842.git.raphaelsc@scylladb.com>
"This patchset allows Scylla to determine the size of a memtable instead
of relying in the user-provided memtable_cleanup_threshold. It does that
by allowing the region_group to specify a soft limit which will trigger
the allocation as early as it is reached.
Given that, we'll keep the memtables in memory for as long as it takes
to reach that limit, regardless of the individual size of any single one
of them. That limit is set to 1/4 of dirty memory. That's the same as
last submission, except this time I have run some experiments to gauge
behavior of that versus 1/2 of dirty memory, which was a preferred
theoretical value.
After that is done, the flush logic is reworked to guarantee that
flushes are not initiated if we already have one memtable under flush.
That allow us to better take advantage of coalescing opportunities with
new requests and prevents the pending memtable explosion that is
ultimately responsible for Issue 1817.
I have run mainly two workloads with this. The first one a local RF=1
workload with large partitions, sized 128kB and 100 threads. The results
are:
Before:
op rate : 632 [WRITE:632]
partition rate : 632 [WRITE:632]
row rate : 632 [WRITE:632]
latency mean : 157.8 [WRITE:157.8]
latency median : 115.5 [WRITE:115.5]
latency 95th percentile : 486.7 [WRITE:486.7]
latency 99th percentile : 534.8 [WRITE:534.8]
latency 99.9th percentile : 599.0 [WRITE:599.0]
latency max : 722.6 [WRITE:722.6]
Total partitions : 189667 [WRITE:189667]
Total errors : 0 [WRITE:0]
total gc count : 0
total gc mb : 0
total gc time (s) : 0
avg gc time(ms) : NaN
stdev gc time(ms) : 0
Total operation time : 00:05:00
END
After:
op rate : 951 [WRITE:951]
partition rate : 951 [WRITE:951]
row rate : 951 [WRITE:951]
latency mean : 104.8 [WRITE:104.8]
latency median : 102.5 [WRITE:102.5]
latency 95th percentile : 155.8 [WRITE:155.8]
latency 99th percentile : 177.8 [WRITE:177.8]
latency 99.9th percentile : 686.4 [WRITE:686.4]
latency max : 1081.4 [WRITE:1081.4]
Total partitions : 285324 [WRITE:285324]
Total errors : 0 [WRITE:0]
total gc count : 0
total gc mb : 0
total gc time (s) : 0
avg gc time(ms) : NaN
stdev gc time(ms) : 0
Total operation time : 00:05:00
END
The other workload was the workload described in #1817. And the result
is that we now have a load that is very stable around 100k ops/s and
hardly any timeouts, instead of the 1.4 baseline of wild variations
around 100k ops/s and lots of timeouts, or the deep reduction of
1.5-rc1."
* 'issue-1817-v4' of github.com:glommer/scylla:
database: rework memtable flush logic
get rid of max_memtable_size
pass a region to dirty_memory_manager accounting API
memtable: add a method to expose the region_group
logalloc: allow region group reclaimer to specify a soft limit
database: remove outdated comment
database: uphold virtual dirty for system tables.
If we're talking to just one replica, the digest is not going to be used,
so better not to calculate it at all. The optimization helps with
LOCAL_ONE queries where the result is large, but does not contain large
blobs (many small rows).
This patch adds a digest_algorithm parameter to the READ_DATA verb that
can take on two values: none and MD5 (default), and sets it to none when
we're reading from one replica.
In the future we may add other values for more hardware-friendly digest
algorithms.
Message-Id: <1479380600-19206-1-git-send-email-avi@scylladb.com>
Currenlty we make the mutation readers for streaming at different
time point, i.e.,
do_for_each(_ranges.begin(), _ranges.end(), [] (auto range) {
make a mutation reader for this range
read mutations from the reader and send
})
If there are write workload in the background, we will stream extra
data, since the later the reader is made the more data we need to send.
Fix it by making all the readers before starting to stream.
Fixes#1815
Message-Id: <1479341474-1364-2-git-send-email-asias@scylladb.com>
If sstable Summary is not present Scylla does not refuses to boot but
instead creates summary information on the fly. There is a bug in this
code though. Summary files is a map between keys and offsets into Index
file, but the code creates map between keys and Data file offsets
instead. Fix it by keeping offset of an index entry in index_entry
structure and use it during Summary file creation.
Reviewed-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20161116165421.GA22296@scylladb.com>
The way we currently flush memtables, we seal the current one but wait
on a semaphore for the actual flush to proceed.
This is pointless, because if the flush is not proceeding we'll use up
memory for the new entries anyway, be them in a newly opened memtable or
not. As a matter of fact, by opening a new memtable we are foregoing
coalescing opportunities.
After recent changes to the flush paths, we are now in a position to do
differently. We move the semaphore earlier, and if we can't acquire it
we keep appending to the current memtable.
For explicit flushes, we'll queue and prioritize them over memory-based
flushes. This has the nice property of potentially coalescing various
flushes for the same CF into one.
Coalescing flushes for the same CF is particularly helpful for
commitlog-initiated flushes that can't complete within the flush period.
What we see currently, is that under heavy load the commitlog will keep
sealing memtables adding to the existing load.
Another interesting property of this approach is that we can keep the
disk utilization higher, by allowing a new flush to start before the
memtable is fully sealed. By design, every time a memtable is finished
flushing it will call revert_potentially_cleaned_up_memory() to revert
the virtual memory charges. That is the perfect moment for us to act.
It indicates that all the data flushing part is done.
The way we'll do it is by keeping the semaphore_units alive for this
memtable. When the flush ends, we destroy that object. This will
effectively trigger the next flush if there is a next flush that can be
initiated.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
After recent changes to the memtable code, there is no reason for us to
uphold a maximum memtable size. Now that we only flush one memtable at a
time anyway, and also have soft limit notifications from the
region_group_reclaimer, we can just set the soft limit to the target
size and let all of that be handled by the dirty_memory_manager.
It does have the added property that we'll be flushing when we globally
reach the soft limit threshold. In conditions in which we have multiple
CF writes fighting for memory, that guarantees that we will start
flushing much earlier than the hard limit.
The threshold is set to 1/4 of dirty memory. While in theory we would
prefer the memtables to go as big as 1/2 of dirty memory, in my
experiments I have found 1/4 to be a better fit, at least for the
moment.
The reason for such behavior is that in situations where we have slow
disks, setting the soft limit to 1/2 of dirty will put us in a situation
in which we may not have finished writing down the memtable when we hit
the limit, and then throttle. When set the threshold to 1/4 of dirty, we
don't throttle at all.
This behavior could potentially be fixed by not doing the full
memtable-based throttling after we do the commitlog throttling, but that
is not something realistic for the moment.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We would like to know from which region is a particular flush coming
from, and account accordingly. The reasoning behind that, is that soon
we'll be driving the flushes internally from the dirty_memory_manager
without explcitly triggering them.
We need to start a flush before the current one finishes, otherwise
we'll have a period without significant disk activity when the current
SSTable is being sealed, the caches are being updated, etc.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
That is technically not needed because a memtable inherits from group. So
whenever we have a memtable, we can use it's group() method to obtain a
group for it, and then from there go to the region_group.
However, region() is a const method in the memtable, so we have to play
trick with the const_cast, or remove the constness from the region. An
alternative to that, which I prefer, is to expose a method for the
region_group directly from the memtable object that does the right thing
and bypasses all that.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
The region_group_reclaimer will let us know every time we are over the
limit we have specified for memory usage.
However, For some applications, we would be interested in knowing about
memory build up earlier, so we can start doing something about it before
we reach that condition.
This patch introduce soft limit notifications for the
region_group_reclaimer. After this patch is applied, start_reclaim() is
called earlier, and stop_reclaim() later, after the soft condition is
abated.
There are methods that allow one to easily test if the pressure
condition is a soft limit condition or a hard, threshold condition and
act accordingly. Whether to act on both conditions or just one of them
is up to the application.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Currently the virtual dirty mechanism is not properly set for system
tables. We haven't divided the system table allowance by two, which
means it won't start thottling earlier as it was supposed to.
In practice, this has little effect because system table requests are
very well behaved, their sizes well known, and they tend to be
force-flushed. But we should be consistent.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
"Cache entries for wide partitions are usually smaller than other
entries and the cost of recreating them is higher so it makes sense
to keep them longer than ordinary entries."
This will allow us to see how big is an amount
of evictions of cached info about wide partitions.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Query pager needs to handle results that contain partitions with
possibly multiple clustering rows quite differently than results with
just one row per partition (for example a page may end in a middle of
partition). However, the logic dealing with partitions with clustering
rows doesn't work correctly for SELECT DISTINCT queries, which are
much more similar to the ones for schemas without clustering key.
The solution is to set _has_clustering_keys to false in case of SELECT
DISTINCT queries regardless of the schema which will make pager
correctly expect each partition to return at most one rows.
Fixes#1822.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1478612486-13421-1-git-send-email-pdziepak@scylladb.com>
Now that the histogram has its own unit expressed in its template
parameter, there is no reason to convert it to nano just so we may need
to convert it back if the histogram needs another unit.
This patch will keep everything as a duration until last moment, and
then we'll convert when needed.
This was suggested by Amnon.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <218efa83e1c4ddc6806c51913d4e5f82dc6d231e.1479139020.git.glauber@scylladb.com>
After the conversion to nonwrapping ranges, construct_range_to_endpoint_map()
may be called with semi-infinite token ranges, but it does not expect this,
calling nonwrapping_range::end()->value() unconditionally.
Fix by checking whether this is a semi-infinite range on the right, and
replace ->value() by maximum_token() instead.
Fixes `nodetool describering` (once more).
Message-Id: <1478983010-29630-1-git-send-email-avi@scylladb.com>
Exception handling was broken because after io checker, storage_io_error
exception is wrapped around system error exceptions. Also the message
when handling exception wasn't precise enough for all cases. For example,
lack of permission to write to existing data directory.
Fixes#883.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <b2dc75010a06f16ab1b676ce905ae12e930a700a.1478542388.git.raphaelsc@scylladb.com>
"JMX metrics were found to be either not showing, or showing absurd
values. Turns out there were multiple things wrong with them. The
patches were sent separately but conflict with one another. This series
is a collection of the patches needed to fix the issues we saw.
Fixes#1832, #1836, #1837"
We are tracking latencies in microseconds, but almost everywhere else
they are reported in microseconds. Instead of just converting, this
patch tries to be a bit more future proof and embed the unit into the
type - and we then default to microseconds.
I have verified that the JMX measures now report sane values for both
the storage proxy and the column family. nodetool cfhistograms still
works fine. That one is reported in nanoseconds, but through the
estimated_histogram, not ihistogram.
Fixes#1836
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We have recently fixed a bug due to which the constructor parameters for
moving average were inverted, leading to the numbers being just plain
wrong. However, the calculation of alpha was already inverted, meaning
it was right by accident and now that's wrong.
With the wrong alpha, the values we see are still correct, but they move
very quickly. The intention of this code is obviously to smooth things
out.
This was found out by Nadav. I have tested and confirmed that the smoothing
factor now works as expected.
Fixes #1837
Signed-off-by: Glauber Costa <glauber@scylladb.com>
moving_averages constructor is defined like this:
moving_average(latency_counter::duration interval, latency_counter::duration tick_interval)
But when it is time to initialize them, we do this:
... {tick_interval(), std::chrono::minutes(1)} ...
As it can be seen, the interval and tick interval are inverted. This
leads to the metrics being assigned bogus values.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <d83f09eed20ea2ea007d120544a003b2e0099732.1478798595.git.glauber@scylladb.com>
Snapshot destructor may free some objects managed by the LSA. That's why
partition_snapshot_reader destructor explicitly destroys the snapshot it
uses. However, it was possible that exception thrown by _read_section
prevented that from happenning making snapshot destoryed implicitly
without current allocator set to LSA.
Refs #1831.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1478778570-2795-1-git-send-email-pdziepak@scylladb.com>
After 7c28ed, the schemas defined in the test became compressed by
default. This patch changes the test so that it is explicit about
which schemas shouldn't define a compressor.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <1478646530-5558-1-git-send-email-duarte@scylladb.com>
"Mostly small changes/additions to the API calls to match Cv3
requirements/semantics, i.e. updated scylla-jmx can implement required
nodetool etc calls in a working fashion."
The CQL 3.1 documentation specifies that for disabling compression,
users should use an empty string:
ALTER TABLE mytable WITH COMPRESSION = {'sstable_compression': ''};
However, Cassandra also accepts the absence of the sstable_compression
option to disable compression. The patch 7c28ed prevented this behavior
in Scylla, which this patch aims to fix.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Message-Id: <1478639499-4183-1-git-send-email-duarte@scylladb.com>
Current speculation target selection logic has several bugs in multi-dc
setup. It may select a non local target for CL=LOCAL and it may select
more than one target to speculate, one of which is non local.
Examples:
1. Two dataceneters: DC1 RF 2, DC2 RF 2 and read with LOCAL_QUORUM.
In this scenario db::filter_for_query() will return both replicas from
local DC and speculation target selection logic will peek one one which
will be in different DC.
2. Two dataceneters: DC1 RF 2, DC2 RF 2 and read with LOCAL_ONE + RRD.DC_LOCAL
In this scenario db::filter_for_query() will return all nodes in local DC and
there already be enough nodes to speculate, but current logic will add
one node from non local dc as a speculation target.
The patch below fixed both of those scenarios.
Message-Id: <20161103154637.GS7766@scylladb.com>
Since continuity flag introduction row cache contains a single dummy
entry. cache_tracker knows nothing about it so that it doesn't appear in
any of the metrics. However, cache destructor calls
cache_tracker::on_erase() for every entry in the cache including the
dummy one. This is incorrect since the tracker wasn't informed when the
dummy entry was created.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1478608776-10363-1-git-send-email-pdziepak@scylladb.com>
partition_range passed to row_cache::make_reader
has to be kept alive as long as the resulting reader
is used.
Otherwise weird things start to happen.
This used to work just because of a pure luck.
When I started changing the row_cache implementation
I run into very weird behaviors for this tests.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Message-Id: <2c9e337dbbcf35f4e1394cad043eda10b8c2bd4a.1478602876.git.piotr@scylladb.com>
Commit 8fca1887c2 ("storage_service: fix range wrapping in
describe_ring") fixed incorrect range wrapping code for describe_ring,
but fails when the number of endpoints for a token is greater than one,
because the endpoints are stored in an unordered vector.
Fix by comparing the endpoints in a way that ignores their order.
Message-Id: <1478460826-15923-1-git-send-email-avi@scylladb.com>
Fixes#1775
stream lacks a check "is_open", which is a bummer. We have to both
prevent exception propagation and add a flag of our own to make sure
exceptions in producer code reaches consumer, and does not simply
get lost in the reactor.
Message-Id: <1478508817-18854-1-git-send-email-calle@scylladb.com>
This patchset adds missing properties to the create_view_statement,
such as whether the view is compact or the order of its clustering
columns.
Fixes#1766
"The atomic sstable deletion provides exception safety at the cost of
quadratic behavior in the number of sstables awaiting deletion. This
causes high cpu utilization during startup.
Change the code to avoid quadratic complexity, and add some unit tests.
See #1812."
In order to ensure exception safety, the atomic sstable deletion code
creates a copy of the list of sstables pending deletion, modifies that
copy, and then replaces the original data with the copy. This guarantees
that any exception does not change the data, since the assignment does
not require allocation.
However, it does result in quadratic behavior. During startup, all
sstables are loaded on each shard, and each shard deletes sstables that
are do not have any partitions served by that shard; this results in
almost all sstables being deleted from all shards, with all that work
going to shard 0; the list grows to O(nr sstables), and there are
O((nr sstables) * (nr shards)) operations to perform.
Fix by replacing the copy-modify-assign method with an in-place update,
but one that is designed to only commit changes after all allocations
have been made; in addition, instead of using a list, use a hash table,
removing another source of quadratic behavior.
Fixes#1812 (the quadratic beahvior part).
"Currently, partition range queries are processed in parallel on all
shards. This is inefficient because we are likely to drop the results
from all but one shard, assuming a well-populated column family. We
are multiplying our work by a factor of smp::count.
While this is worthwhile in its own right, it is really an excuse to
sneak in the range/shard generator (patch 5), which is preliminary for
a new sharding algorithm, dividing tokens among shards based on the
middle-significant bits rather than the most-siginificant bits (which
alias with vnodes)
Fixes #1573."
Instead of asking a shard for cmd->partition_limit and cmd->row_limit,
just ask it for the number of partitions and rows still needed to
satisfy the query. This removes the need to trim the shard's result.
Since every shard might cause the row_limit quota to be satisfied, every
shard might be the last one we need. Hence it is better to process shards
sequentially, stopping if the quota is reached or the range is exhausted.
The original code tried to yield to reduce latency, but this is now
unnecessary, as we're doing a lot less work per iteration (if it becomes
necessary, we should do it on the replica shard, not the coordinating shard).
Building on the single-range sharder, add a sharder for vectors of
partition ranges. This helps with wrapped ranges, which are translated
into a vector containing two shards.
Divides a ring_position range into a sequence of shard/range pairs. This
allows sequential iteration over shards in ring order.
The current multi-partition query executes on all shards in parallel, but
this is very wasteful, as most of the data will be thrown away if it is not
included in the page. With the generator, we can switch to sequential
execution.
When performing a range query, we want to iterate over shards, running the
query on each shard in order until the query range is exhausted or we have
the right number of rows.
To be able to do this, introduce token_for_next_shard(), which allows us
to determine the boundary between shards.
It is a sort-of inverse to shard_of(), in that
shard_of(token_for_next_range(t)) == shard_of(t) + 1
- Add a inserts, updates, deletes members to cql_stats.
- Store cql_stats& in a modification_statement and increment the corresponding counter according to the value of a "type" field.
- Store cql_stats& in a batch_statement and increment the statistics for each BATCH member.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
- Add a "reads" counter to a cql3::cql_stats struct.
- Store a reference for a query_processor::_cql_stats in the select_statement object.
- Increment a "reads" counter where needed.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Right now we are calculating latencies only when we are about to add an
item to the memtable.
That's incorrect and misleading, for two reasons. First, it leaves the
commitlog latencies out. But second, it is done after the memtable wall
effect is applied, which means we are not counting throttle time neither
in the memtables or in the commitlog.
To do that, we'll start the latency_counter object as soon as possible
and move it all the way to apply_in_memory(). That should span the
entire write operation.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <4e424780d290fd5938046060df2b17e2b470b717.1478111467.git.glauber@scylladb.com>
There are two variants of apply_in_memory() being called in do_apply():
with and without the commitlog. The main differences are that when the
commitlog is involved, we need to wait for its future to complete before
moving to apply_in_memory. That can easily be factored out by providing
an always-ready future if we don't have the commitlog enabled, and
waiting on that.
The second, is that the commitlog version can cause apply_in_memory to
generate an exception if there is replay position reordering. However,
there is no harm in appending the exception handler to both versions. In
one of them it's an impossible exception, but that's fine.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <8cee0cad9b1930a057a24e095f0a655069ae8be2.1478111467.git.glauber@scylladb.com>
There are places in which we need to use the column family object many
times, with deferring points in between. Because the column family may
have been destroyed in the deferring point, we need to go and find it
again.
If we use lw_shared_ptr, however, we'll be able to at least guarantee
that the object will be alive. Some users will still need to check, if
they want to guarantee that the column family wasn't removed. But others
that only need to make sure we don't access an invalid object will be
able to avoid the cost of re-finding it just fine.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <722bf49e158da77ff509372c2034e5707706e5bf.1478111467.git.glauber@scylladb.com>
Since size estimates are stored as wrapped ranges, we call compat::wrap()
to convert from the now-standard unwrapped ranges back to wrapped ranges.
However, compat::wrap() relies on the ranges being in sorted order,
but our input is not. This leads to a crash as we find an unexpected
empty token in the middle of the vector.
Sort it so compat::wrap() works as expected.
Fixes#1804.
Message-Id: <1478161908-25051-1-git-send-email-avi@scylladb.com>
Wrapping ranges are a pain, so we are moving wrap handling to the edges.
Since cql can't generate wrapping ranges, this means thrift and the ring
maintenance code; also range->ring transformations need to merge the first
and last ranges.
Message-Id: <1478105905-31613-1-git-send-email-avi@scylladb.com>
We were ignored --smp option taken from io.conf since iotune didn't supported
it, but now it supported we can pass it.
(We need to pass it because we need to measure io performance on same condition
with scylla)
Fixes#1768
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1478082591-27205-1-git-send-email-syuu@scylladb.com>
Under the hood, the selectable::add_and_get_index() function
deliberately filters out duplicate columns. This causes
simple_selector::get_output_row() to return a row with all duplicate
columns filtered out, which triggers and assertion because of row
mismatch with metadata (which contains the duplicate columns).
The fix is rather simple: just make selection::from_selectors() use
selection_with_processing if the number of selectors and column
definitions doesn't match -- like Apache Cassandra does.
Fixes#1367
Message-Id: <1477989740-6485-1-git-send-email-penberg@scylladb.com>
This patch adds an `update-version` script for updating the Scylla
version number in `SCYLLA-VERSION-GEN` file and committing the change to
git.
Example use:
$ ./scripts/update-version 1.4.0
which results into the following git commit:
commit 4599c16d9292d8d9299b40a3e44ef7ee80e3c3cf
Author: Pekka Enberg <penberg@scylladb.com>
Date: Fri Oct 28 10:24:52 2016 +0300
release: prepare for 1.4.0
diff --git a/SCYLLA-VERSION-GEN b/SCYLLA-VERSION-GEN
index 753c982..eba2da4 100755
--- a/SCYLLA-VERSION-GEN
+++ b/SCYLLA-VERSION-GEN
@@ -1,6 +1,6 @@
#!/bin/sh
-VERSION=666.development
+VERSION=1.4.0
if test -f version
then
Message-Id: <1477639560-10896-1-git-send-email-penberg@scylladb.com>
Fixes#1709.
* 'refresh-resilient-v3' of github.com:raphaelsc/scylla:
db: make refresh resilient to permission denied error
db: make it possible to use custom error handler with io checker
sstables: remove duplicated declaration of remove_by_toc_name
We use `data_resource` class in the CQL parser, which let's users refer
to a table resource without specifying a keyspace. This asserts out in
get_level() for no good reason as we already know the intented level
based on the constructor. Therefore, change `data_resource` to track the
level like upstream Cassandra does and use that.
Fixes#1790
Message-Id: <1477599169-2945-1-git-send-email-penberg@scylladb.com>
We store all auth perm strings in upper case, but the user might very
well pass this in upper case.
We could use a standard key comparator / hash here, but since the
strings tend to be small, the new sstring will likely be allocated in
the stack here and this approach yields significantly less code.
Fixes#1791.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <51df92451e6e0a6325a005c19c95eaa55270da61.1477594199.git.glauber@scylladb.com>
User may forget to set permission of new sstables in upload dir
before refreshing them, and that will result in shutdown.
io_checker is now able to work with a custom handler, so all we
have to do is to whitelist EACCES.
Fixes#1709.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
By default, io checker will cause Scylla to shutdown if it finds
specific system errors. Right now, io checker isn't flexible
enough to allow a specialized handler. For example, we don't want
to Scylla to shutdown if there's an permission problem when
uploading new files from upload dir. This desired flexibility is
made possible here by allowing a handler parameter to io check
functions and also changing existing code to take advantage of it.
That's a step towards fixing #1709.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
EC2 paravirtual instances uses pv-grub, which refers /boot/grub/menu.lst (grub0.9x config file) instead of grub2 config file.
So add boot parameters on /boot/grub/menu.lst when the file exists, and the instance is on EC2.
Fixes#1598
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1472056875-17512-1-git-send-email-syuu@scylladb.com>
wget is often used from scripts recording to logs; as it emits a log
line every second, the logs are huge and unreadable. Make it quieter.
Message-Id: <1477558534-32718-1-git-send-email-avi@scylladb.com>
"5ff699e09fcbd62611e78b9de601f6c8636ab2f0 ("row_cache: rework cache to
use fast forwarding reader") brought some significant changes to the
row cache implementation. Unfortunately, "significant changes" often
translates to "more bugs" and this time was no different.
This series contains fixes for the problems introduced in that rework
and makes failing dtest
bootstrap_test.py:TestBootstrap.local_quorum_bootstrap_test
pass again."
* 'pdziepak/cache-fixes/v1' of github.com:cloudius-systems/seastar-dev:
row_cache: avoid dereferencing invalid iterator
row_cache: set _first_element flag correctly
row_cache: fix clearing continuity flag at eviction
Conditions in row_cache::do_find_or_create_entry() make it possible that
std::prev(it) is going to be dereferenced even if it is a begin
iterator.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
If the continuity flag was set for the first element _first_element flag
would not be cleared. This shouldn't cause any correctness problems but
properly setting the flag allows to avoid some unnecessary key
comparisons.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
In original implementation the continuity flag indicated that cache has
full information about the range the between current partition and the
one following it, hence when evicting an entry the one preceeding it
had to have its continuity flag cleared.
This was changed, however, and now the continuiy flag tells whether the
cache is continuous between the current element and the one before it.
This means that eviction code needs to clear the flag for the entry
directly following the evicted one.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
* seastar 69acec1...98b5a2d (9):
> rpc: Silence warning about ignored failed future
> future: prioritise continuations that can run immediately
> iotune: relax aio restrictions
> build: support for static linking with boost
> rpc: Fix crash during connection teardown
> rpc: Move _connected flag to protocol::connection
> rpc test: fail test if exception is thrown during test execution
> rpc: do not assume underling semaphore type
> rpc: fix default resource limit
In boost 1.60, the executable's command-line arguments are expected to
be separated from the boost command-line arguments by '--'. Detect
this requirement and comply with it.
Message-Id: <1477212424-3831-1-git-send-email-avi@scylladb.com>
We calculate two sizes during the allocation: "size", which is the
in-segment size of this mutation, and "s", which is that plus the
overhead. cycle() must be called with the latter, not the former, as
doing otherwise may lead to buffer overflows.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <ccf346d8d0ebb44a1ba9fd069653bab0d7be0a61.1477063157.git.glauber@scylladb.com>
"This patchset reworks the commitlog logic to better handle conditions in
which we are getting requests faster than the disk can handle. It does
this by building a wall around the commitlog and only allowing
allocations to proceed when we are under the desired memory threshold.
The main advantage of that is that we can now easily set the commitlog
to work at disk speed, more or less allowing an "one byte in for each
byte out" approach instead of depending on the current cycle to finish.
As a result, max latencies are greatly reduced.
Testing Results
===============
To test this, I have ran a workload that times out frequently. That
workload use 10 threads to write 100 partitions (to isolate from the
effects of the memtable introduced latencies) in a loop and each
partition is 2MB in size.
After 10 minutes running this load, we are left with the following
percentiles:
latency mean : 51.9 [WRITE:51.9]
latency median : 9.8 [WRITE:9.8]
latency 95th percentile : 125.6 [WRITE:125.6]
latency 99th percentile : 1184.0 [WRITE:1184.0]
latency 99.9th percentile : 1991.2 [WRITE:1991.2]
latency max : 2338.2 [WRITE:2338.2]
After this patch:
latency mean : 54.9 [WRITE:54.9]
latency median : 43.5 [WRITE:43.5]
latency 95th percentile : 126.9 [WRITE:126.9]
latency 99th percentile : 253.9 [WRITE:253.9]
latency 99.9th percentile : 364.6 [WRITE:364.6]
latency max : 471.4 [WRITE:471.4]
I have run this with larger sizes as well, and it generally performs
much better than the baseline version. For sizes up to 5MB, I have seen
no timeouts in my setup. After that, I see some timeouts. Buffer
splitting is expected to make this better.
Aside from performance testing, this was also tested with batch and
periodic mode for various requests sizes."
Current tracker for pending allocations is a queue_size GAUGE. Add a
total_operations version so we have more insight on what's going on.
It will be called requests_blocked_memory for consistency with other
subsystems that track similar things.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
scylla-blocktune is a script that parses scylla.yaml and tunes the data file
and commitlog directories it references.
Tuning includes:
- set the I/O scheduler to noop
- disable merging
- tune dependent devices (like RAID members)
Message-Id: <1476357027-15014-2-git-send-email-avi@scylladb.com>
The current chunk size of 256 gives a 50% probability of a 128k read or
write getting split into two accesses. This reduces efficiency and
increases latency.
Change the chunk size to 1MB, with a 12% probability of cross-member
access.
Message-Id: <1476269082-2473-1-git-send-email-avi@scylladb.com>
Since Debian 8(jessie) does not provides g++-5, we frequently got compile error
because we are using older compiler.
To fix the problem, backport g++-5 from Debian 9(stretch).
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1476694318-10640-3-git-send-email-syuu@scylladb.com>
On Debian, lsb_release -r returns the version number something like '8.6'.
However, on this script we want to check major version only.
Therefore we can use VERSION_ID from /etc/os-release which only contains
major version number.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1476694318-10640-2-git-send-email-syuu@scylladb.com>
Commit 7dcd70124a "tests/sstables: add
test for fast forwarding reader" added a test for skipping parts of
sstable. Unfortunately, it did not include the sstables it was trying to
read.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
The current incarnation of commitlog establishes a maximum amount of
writes that can be in-flight, and blocks new requests after that limit
is reached.
That is obviously something we must do, but the current approach to it
is problematic for two main reasons:
1) It forces the requests that trigger a write to wait on the current
write to finish. That is excessive; ideally we would wait for one
particular write to finish, not necessarily the current one. That
is made worse by the fact that when a write is followed by a flush
(happens when we move to a new segment), then we must wait for
*all* writes in that segment to finish.
1) it casts concurrency in terms of writes instead of memory, which
makes the aforementioned problem a lot worse: if we have very big
buffers in flight and we must wait for them to finish, that can
take a long time, often in the order of seconds, causing timeouts.
The approach taken by this patch is to replace the _write_semaphore
with a request_controller. This data structure will account the amount
of memory used by the buffers and set a limit on it. New allocations
will be held until we go below that limit, and will be released
as soon as this happens.
This guarantees that the latencies introduced by this mechanism are
spread out a lot better among requests and will keep higher percentile
latencies in check.
To test this, I have ran a workload that times out frequently. That
workload use 10 threads to write 100 partitions (to isolate from the
effects of the memtable introduced latencies) in a loop and each
partition is 2MB in size.
After 10 minutes running this load, we are left with the following
percentiles:
latency mean : 51.9 [WRITE:51.9]
latency median : 9.8 [WRITE:9.8]
latency 95th percentile : 125.6 [WRITE:125.6]
latency 99th percentile : 1184.0 [WRITE:1184.0]
latency 99.9th percentile : 1991.2 [WRITE:1991.2]
latency max : 2338.2 [WRITE:2338.2]
After this patch:
latency mean : 54.9 [WRITE:54.9]
latency median : 43.5 [WRITE:43.5]
latency 95th percentile : 126.9 [WRITE:126.9]
latency 99th percentile : 253.9 [WRITE:253.9]
latency 99.9th percentile : 364.6 [WRITE:364.6]
latency max : 471.4 [WRITE:471.4]
Signed-off-by: Glauber Costa <glauber@scylladb.com>
In a subsequent patch, I'll use this code in a different place. To
prepare for that, we move it out as a method. It also fits a lot better
inside the segment manager, so move it there.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Goal is to calculate a size that is lesser or equal than the
segment-dependent size.
This was originally written by Tomasz, and featured in his submission
"commitlog: Handle overload more gracefully"
Extracted here so it sits clearly in a different patch.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
It is mostly an optimization, and while it makes sense in this context,
it won't soon as we'll stop waiting for the current cycle specifically
to finish.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We'll do that so we can, in following patches, use static members from
the segment. Those are not defined at this point.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We track the amount of pending allocations but we don't really export
it. It will be crucial when we stop tracking pending writes.
This patch exports it through a method instead of the totals structure,
so we can easily change it. Current code probing pending_allocations
(the api code) is also converted to use the public method instead of the
totals struct.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
"This patchset enables mutation readers to be fast forwarded to a different
partition range. The main reason for introducing such feature are range
queries served from cache. If the cache is partially populated in the
requested range the reader will end up with multiple subranges that have
to be read from the sstables. Originally, each of these subranges would
require a new reader to be created, but with fast forwarding we can have
just one sstable reader. This is better since there is a chance that buffers
kept by the reader may be still useful after fast forwarding it.
In this series there are also patches that clean up cache readers in order
to make integration with fast forwarding easier. Namely, continuity flag is
changed to store information about range before the entry which significantly
simplifies the logic.
Fixes #1299."
* 'pdziepak/fast-forward-mutation-readers/v5' of github.com:cloudius-systems/seastar-dev: (24 commits)
sstables: keep separate stream history for single and range reads
sstables: drop sstable::{lower, upper}_bound()
row_cache: rework cache to use fast forwarding reader
row_cache: put cache entry flags in a struct
row_cache: add do_find_or_create_entry() to reduce code duplication
mutation_reader: forward fast_forward_to() calls
tests/row_cache: add fast_forward_to() to throttled reader
tests/row_cache: count mutations read from _underlying
memtable: add support for fast_forward_to()
drop key readers
tests/mutation_reader: test fast forwarding combined reader
database: enable fast forwarding of range_sstable_reader
combined_mutation_reader: implement fast_forward_to()
mutation_reader: make combinded_reader public
tests/sstables: add test for fast forwarding reader
tests: add more helpers to mutation reader assertions
sstables: enable fast forwarding for range readers
mutation_reader: introduce fast_forward_to()
sstables: implement mutation_reader::impl::fast_forward_to()
sstables: introduce index_reader
...
Single partition and partition range reads are expected to behave
considerably different so it is worth to have them use separate file
stream history. This also makes reads use different history for each
sstable which is also a good thing.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
This uncomfortably large patch overhauls cache range reader so that it
can take advantage of fast forwarding mutation readers.
A significant change in the cache itself is that the continuity flag now
is used to determine whether cache is contiguous between the previous
entry and the current one. This allows for a significant simplification
of the cache code and easier integration with reader fast forwarding.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Flags are easier to manage if they are in a single structure.
Especially, default initialization and move contstructors are simpler
and less error prone.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Originally, cache tests checked how many times a mutation reader was
created from the underlying mutation source to determine whether
continuity flag is working correctly.
This is not going to work with fast forwarding mutation readers so the
test is switched to count number of mutations (+ end of stream markers)
returned from underlying mutaiton readers which is much less fragile.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Fast forwarding of memtable readers is needed only for unit tests which
often use memtables as underlying data source for cache and the cache
readers.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
When fast forwarding a reader that combines sstable reader we must also
remember that the set of sstables for the new range may be different
than for the previous one. The reader introduced in this patch makes
sure that we read from correct sstables.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
We want to be able to fast forward sstable readers. However, just
implementing fast_forward_to() for combined_reader is not enough as the
sstables we are reading from may need to change.
Following patches are going to introduce a combined sstable reader that
derives from combined_reader. To make that possible we first need to
make combined_reader public.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
This patch introduces the interface for fast forwarding mutation
readers. The main user of this feature is going to be cache which, while
serving range query, may need to read multiple small ranges from the
sstables to populate itself with the missing entries.
Fast forwarding is an alternative to recreating a reader with different
range. Its main advantage is fact that it avoids dropping data that has
already been read.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
This patch allows sstable readers to be fast forwarded without making it
necessary to recreate the reader (and dropping all buffers in the
process). It is built on top of index_reader and ability of
data_consume_context to be fast forwarded.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
index_reader is a helper that implements index lookups. Its goal is to
avoid dropping read buffers if they still may be needed (for example to
get end bound of the range or after fast forwarding the reader).
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
That overload was used only by unit test and violated guarantee that
partition range lives until mutation reader is done.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
is_compactible() will pass on very small regions. full_compaction() is
only used in tests to force objects to be moved due to compaction, so
we want all reclaimable regions to be compacted.
The move constructor of partition_version was not invoking move
constructor of anchorless_list_base_hook. As a result, when
partition_version objects were moved, e.g. during LSA compaction, they
were unlinked from their lists.
This can make readers return invalid data, because not all versions
will be reachable.
It also casues leaks of the versions which are not directly attached
to memtable entry. This will trigger assertion failure in LSA region
destructor. This assetion triggers with row cache disabled. With cache
enabled (default) all segments are merged into the cache region, which
currently is not destroyed on shutdown, so this problem would go
unnoticed. With cache disabled, memtable region is destroyed after
memtable is flushed and after all readers stop using that memtable.
Fixes#1753.
Message-Id: <1476778472-5711-1-git-send-email-tgrabiec@scylladb.com>
This patch uses cf_properties instead to add the missing attributes to
the create_view_statement class.
Fixes#1766
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds the VIEWS element to the cause enum so we can
mark failures due to incomplete support of materialized views.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch extracts the definition of the default compressor into the
compression_parameters class, so that the table and view creation
statements don't have to explicitly deal with it.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch extracts the cf_properties class, which contains common
attributes of tables and materialized views.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Since we are exiting Scylla process in engine().at_exit() using
::_exit(0), even verify_seastar_io_scheduler() throwing an exception,
scylla always exit with 0.
Systemd misunderstands scylla-server.service was shutdown successfully
because of this, so we need to pass correct exit code to ::_exit() here.
Fixes#1674
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1475065607-15486-1-git-send-email-syuu@scylladb.com>
* seastar 207bf3d...ccd8649 (3):
> Merge "Augment semaphore with non-blocking operations" from Glauber
> Merge "More dynamic fstream patches" from Paweł
> Merge "fstream: add dynamic adjustments based on stream history" from Paweł
A 1MB response will require 2000 allocations with the current 512-byte
chunk size. Increase it exponentially to reduce allocation count for
larger responses (still respecting the upper limit).
Message-Id: <1476369152-1245-1-git-send-email-avi@scylladb.com>
Memory accounting code was attaching partition_snapshot to
partition_entry in order to calculate the size of partition_version
object. However, it is only allowed if partition_entry doesn't have
any snapshot attached already. In this case it always has one, created
by the flushing reader.
Change the accounting code to reuse existing partition_snapshot reference.
Fixes#1746
Message-Id: <1476449160-9252-1-git-send-email-tgrabiec@scylladb.com>
LSA tries to allocate zones as large as possible (while still leaving
enough free space for the standard allocator). It uses the amount of
free memory in order to guess how much it can get, but that obviously
doesn't account for fragmentation and the allocation attempt may fail.
This patch changes the LSA code so that it doesn't throw in case zone
couldn't be created but just returns a null pointer which should be
more performant if the LSA memory cannot grow any more.
Fixes#1394.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1476435031-5601-1-git-send-email-pdziepak@scylladb.com>
The expected behaviour in the scylla_setup script is that a question
will be followed by the answer.
For example, after asking if the scylla should be run as a service the
relevant actions will be taken before the following question.
This patch address two such mis-orders:
1. the scylla-housekeeping depends on the scylla-server, but the
setup should first setup the scylla-server service and only then ask
(and install if needed) the scylla-housekeeping.
2. The node_exporter should be placed after the io_setup is done.
Fixes#1739
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1476370098-25617-1-git-send-email-amnon@scylladb.com>
Change abstract_replication_strategy::create_replication_strategy() to
throw exceptions::configuration_error if replication strategy class
lookup to make sure the error is converted to the correct CQL response.
Fixes#1755
Message-Id: <1476361262-28723-1-git-send-email-penberg@scylladb.com>
* seastar f937fb0...207bf3d (11):
> Merge "iotune: gracefully exit on predictable exceptions" (Fixes#1623)
> core/semaphore: Add semaphore_units::release()
> Merge "rometheus API with grafana uses labels" from Amnon
> core/thread: Fix stack alloc-dealloc mismatch
> core/thread: Make jmp_buf_link::yield_at use the same time point as thread_scheduling_group
> file: support for XFS on older kernels
> reactor: fix bug when handling EBADF in flush_pending_aio()
> prometheus CPU should start in 0
> Collectd: bytes ordering depends on the type
> tests: Check that backtrace() doesn't corrupt signal mask
> core/thread: Add stack guards to seastar thread stacks
If we have a range query involving a wrapping range (i.e., from thrift),
and mutations from both halves of the result are involved, then
we will return the results in the wrong order (and potentially the wrong
partitions) since we order by token, so the results from the second half
of the wrapping range end up before the first.
Fix by splitting the two queries, and merging the second half with lower
priority compared to the first half.
Note: this will be fixed in a better way once we have the sharding iterator,
as then we can query sequentially.
Fixes#1761.
Message-Id: <1476262693-30162-1-git-send-email-avi@scylladb.com>
"This series address two issues that interfere with running the node_exporter as a service in ubuntu 16.
1. The service file should be packed in the deb file
2. When setting the node_exporter as a service it doesn't need to run with scylla use"
* 'amnon/node_exporter_ubuntu_v2' of github.com:cloudius-systems/seastar-dev:
node-exporter service: No need to run as scylla user
debian package: Include the node_exporter service file
the node-exporter does not need to run as scylla user. It can run
without scylla or without the scylla user being configure.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
"This patch-set re-implements the describe_splits_ex() verb to more closely
follow Cassandra's implementation, on which some clients rely.
Ref #1139
Ref #693"
* 'describe-splits/v2' of github.com:duarten/scylla:
thrift: Implement describe_splits_ex based on Cassandra
storage_service: Implement get_splits() function
sstables: Add function to get key samples
sstables/key: Add to_partition_key function
size_estimates_recorder: Increase estimate accuracy
sstables: Get estimates for a particular range
sstables/key: Make key::kind public
The script mistakenly split value at "," when cpuset list is separated
by comma. Instead of matching possible patterns of the argument, let's
pass all characters until reach to space delimiter or end of line.
Fixes#1716
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1476171037-32373-1-git-send-email-syuu@scylladb.com>
This patch re-implements the describe_splits_ex() verb to more closely
follow Cassandra's implementation, on which some clients rely.
Ref #1139
Ref #693
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch implements the get_splits() function in storage_service,
used to split a particular token range in slices of approximately the
specified size, using the sample keys and estimates of the CF's
sstables.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch implements the get_key_samples() function, on which a
future patch will base an implementation of the describe_splits()
thrift verb closer to Cassandra's.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds the estimated_keys_for_range() function, which
estimates the number of keys present between the specified range.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
"The version is taken from the installation rather than the API, a mode command
line indicated that this is part of the setup and uuid is used for the
interaction with the checkversion server."
* 'amnon/check_version_on_startup_v3' of github.com:cloudius-systems/seastar-dev:
scylla_setup: Check and report the scylla version
scylla-housekeeping: check version during setup
There is already queue_length-requests_blocked_memory, but it's a
gauge so does not reflect what happened between the sampling points.
total_operations-requests_blocked_memory will allow to see if there
were any (and how many) requests which were blocked by dirty memory.
Message-Id: <1476098616-12682-1-git-send-email-tgrabiec@scylladb.com>
Presents current heap profile recording.
Works in text mode or dumps to collapsed stacks format from which
flame graph can be generated.
To generate a flamegraph:
(gdb) scylla heapprof --flame
Wrote heapprof.stacks
$ flamegraph.pl --colors mem < heapprof.stacks > heapprof.svg
flamegraph.pl comes from:
https://github.com/brendangregg/FlameGraph.git
Text mode example:
(gdb) scylla heapprof --min 100000000
All (274699676, #10213)
\-- void* memory::cpu_pages::allocate_large_and_trim<memory::cpu_pages::allocate_large_aligned(unsigned int, unsigned int)::{lambda(unsigned int, unsigned int)#1}>(unsigned int, memory::cpu_pages::allocate_large_aligned(unsigned int, unsigned int)::{lambda(unsigned int, unsigned int)#1}) + 169 (268435456, #1)
memory::allocate_large_aligned(unsigned long, unsigned long) + 87
memory::allocate_aligned(unsigned long, unsigned long) + 48
aligned_alloc + 9
logalloc::segment_zone::segment_zone() + 304
logalloc::segment_pool::allocate_segment() + 477
logalloc::segment_pool::segment_pool() + 304
__tls_init.part.801 + 72
logalloc::region_group::release_requests() + 1333
logalloc::region_group::add(logalloc::region_group*) + 514
The branches are formatted like this:
-- <symbol> (<size>, #<count>)
Where <size> is total size of live objects and <count> is total
number of live objects, for all objects allocated from paths going
through this node.
Nodes which share the same <size> and <count> are stacked like this:
-- <symbol_1> (<size>, #<count>)
<symbol_2>
<symbol_3>
Message-Id: <1475583334-19524-1-git-send-email-tgrabiec@scylladb.com>
Limiting the concurrency of memtable flushes to 4 was a temporary
workaround for the fact that we lacked good write behind support. Now
that write behind is properly merged we can reduce the concurrency to
what it should be, one.
This means that memtable flushes will now be serialized, and only when
one of them ends will the next one begin. Disk parallelism is obtained
through the write-behind mechanism.
Fixes#1373
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <528f9ef928b5101bed952df600eb8555c275497a.1475881100.git.glauber@scylladb.com>
There is a limit to concurrency of sstable readers on each shard. When
this limit is exhausted (currently 100 readers) readers queue. There
is a timeout after which queued readers are failed, equal to
read_request_timeout_in_ms (5s by default). The reason we have the
timeout here is primarily because the readers created for the purpose
of serving a CQL request no longer need to execute after waiting
longer than read_request_timeout_in_ms. The coordinator no longer
waits for the result so there is no point in proceeding with the read.
This timeout should not apply for readers created for streaming. The
streaming client currently times out after 10 minutes, so we could
wait at least that long. Timing out sooner makes streaming unreliable,
which under high load may prevent streaming from completing.
The change sets no timeout for streaming readers at replica level,
similarly as we do for system tables readers.
Fixes#1741.
Message-Id: <1475840678-25606-1-git-send-email-tgrabiec@scylladb.com>
Make split_after() more generic by allowing split_point to be anywhere,
not just within the input range. If the split_point is before, the entire
range is returned; and if it is after, stdx::nullopt is returned.
"before" and "after" are not well defined for wrap-around ranges, so
but we are phasing them out and soon there will not be
wrapping_range::split_after() users.
This is a prerequisite for converting partition_range and friends to
nonwrapping_range.
Message-Id: <1475765099-10657-1-git-send-email-avi@scylladb.com>
Commit log replay is a synchronous operation in bootstrap, so services
will only be started after it's completed. By starting compaction before,
less bandwidth will be available to both and consequently boot will be
slowed down. Fix is simply about moving compaction, which is an
asynchronous operation after commitlog replay is over.
Fixes#1620.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <d2a173a4ee4d474317b970c6b39530e61067fea9.1475527955.git.raphaelsc@scylladb.com>
This patch adds the parsing for the "CREATE MATERIALIZED VIEW" statement,
following Cassandra 3 syntax. For example:
CREATE MATERIALIZED VIEW building_by_city
AS SELECT * FROM buildings
WHERE city IS NOT NULL
PRIMARY KEY(city, name);
It also adds the "IS NOT NULL" operator needed for this purpose.
As in Cassandra, "IS NOT NULL" can only be used for materialized
view creation, and not in a normal SELECT. It can only be used with
the NULL operand (i.e., "IS NOT 3" will be a syntax error).
The current implementation of this statement just does some sanity
checking (such as to verify that "city" is a valid column name and that
the "building" base table exists), complains that materialized views are
not yet supported:
SyntaxException: <ErrorMessage code=2000 [Syntax error in CQL query] message="Failed parsing statement: [CREATE MATERIALIZED VIEW building_by_city AS
SELECT * FROM buildings
WHERE city IS NOT NULL
PRIMARY KEY(city, name);] reason: unsupported operation: Materialized views not yet supported">
As mentioned above, the "IS NOT NULL" restriction is not allowed in
ordinary selects not creating a materialized views:
SELECT * FROM buildings WHERE city IS NOT NULL;
InvalidRequest: code=2200 [Invalid query] message="restriction 'city IS NOT null' is only supported in materialized view creation"
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <1475742927-30695-1-git-send-email-nyh@scylladb.com>
The latest virtual dirty patches broke the SSTable tests. The reason for
this is that those tests will flush synthetic memtables that do not have
a region_group attached to it.
Normally in cases like this we would just give the flush_reader an empty
region group. However, the memtable class constructor takes a
region_group pointer and that can be null according to the interface.
So we must conditionally test it.
If there isn't a region_group involved, the virtual dirty accounting
should be disabled: after all, we won't even have the baseline memory
to begin with.
One of the approaches to fix this could be to just provide null
accounter classes to be used as a surrogate for the accounting classes
in this case. However, since this is mostly used for tests, a much
simpler way is to just revert back to the scanning reader in that case.
The scanning reader is similar enough to the flush_reader, except that
it can handle partial ranges, slices, and delegate accesses to an
sstable post-flush. We don't need any of that, but as argued above,
there is no need to remove it either.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
Message-Id: <1475667271-60806-1-git-send-email-glommer@scylladb.com>
Remove inclusions from header files (primary offender is fb_utilities.hh)
and introduce new messaging_service_fwd.hh to reduce rebuilds when the
messaging service changes.
Message-Id: <1475584615-22836-1-git-send-email-avi@scylladb.com>
"Description:
============
Scylla currently suffers from a brick wall behavior of the request throttler.
Requests pile up until we reach the dirty memory limit, at which point we stop
serving them until we have freed enough memory to allow for more requests.
The problem is that freeing dirty memory means writing an SSTable to completion.
That can take a long time, even if we are blessed with great disks. Those long
waiting times can and will translate into timeouts. That is bad behavior.
What this patch does is introduce one form of virtual dirty memory accounting.
Instead of allowing 100 % of the dirty memory to be filled up until we stop
accepting requests, we will do that when we reach 50 % of memory. However,
instead of releasing requests only when an SSTable is fully written, we start
releasing them when some memory was written.
The practical effect of that, is that once we reach 50 % occupancy in our dirty
memory region, we will bring the system from CPU speed to disk speed, and will
start accepting requests only at the rate we are able to write memory back.
Results
=======
With this patchset running a load big enough to easily saturate the disk,
(commitlog disabled to highlight the effects of the memtable writer), I am able
to run scylla for many minutes, with timeouts occurring only when I run out of
disk space, whereas without this patch a swarm of timeouts would start merely 2
seconds after the load started - and would never get stable.
In V2, I have sent a set of graphs illustrating the performance of this solution.
This version does not have any significant differences in that front.
For details, please refer to
https://groups.google.com/d/msg/scylladb-dev/iCvD-3Z-QqY/EM8KUh_MAQAJ
Accuracy of the accounting:
---------------------------
It is important for us to be as accurate as possible when accounting freed
memory, since every byte we mark as freed may allow one or more requests to be
executed. I have measured the accuracy of this approach (ignoring padding,
object size for the mutation fragments) to be 99.83 % of used memory in the
test workload I have ran (large, 65k mutations). Memtables under this circumnstance
tend to have a very high occupancy ratio because throttle breeds idle, and idle
breeds compact-on-idle.
Known Issues:
-------------
A lot of time can be elapsed between destroying the flush_reader and actually
releasing memory. The release of memory only happens when the SSTable is fully
sealed, and we have to flush the files, as well as finish writing all SSTable
components at this point. This happened in practice with a buggy kernel that
would result in flushes taking a long time.
After that is fixed, this is just a theoretical problem and in practice it
shouldn't matter given the time we expect those operations to take."
* 'virtual-dirty-v6' of github.com:glommer/scylla:
database: allow virtual dirty memory management
streamed_mutation: make _buffer private
add accounting of memory read to partition_snapshot_reader
move partition_snapshot_reader code to header file
LSA: allow a group to query its own region group
memtables: split scanning reader in two
sstables: use special reader for writing a memtable
LSA: export information about object memory footprint
LSA: export information about size of the throttle queue
database: export virtual dirty bytes region group
* seastar 18f7bb8...f937fb0 (5):
> Merge "Fix signal mask corruption" from Tomasz
> core/memory: Avoid violating strict aliasing when accessing allocation sites
> core/memory: Avoid indirection when storing allocation sites
> core/memory: Add a way to disable abort on allocation failure in some scope
> core/sharded: Allow mapper to take the service by non-const reference
Scylla currently suffers from a brick wall behavior of the request throttler.
Requests pile up until we reach the dirty memory limit, at which point we stop
serving them until we have freed enough memory to allow for more requests.
The problem is that freeing dirty memory means writing an SSTable to completion.
That can take a long time, even if we are blessed with great disks. Those long
waiting times can and will translate into timeouts. That is bad behavior.
What this patch does is introduce one form of virtual dirty memory accounting.
Instead of allowing 100 % of the dirty memory to be filled up until we stop
accepting requests, we will do that when we reach 50 % of memory. However,
instead of releasing requests only when an SSTable is fully written, we start
releasing them when some memory was written.
The practical effect of that is that once we reach 50 % occupancy in our dirty
memory region, we will bring the system from CPU speed to disk speed, and will
start accepting requests only at the rate we are able to write memory back.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
It is currently protected, but now all users go through
push_mutation_fragment(). So we can safely move its visibility to guarantee
that it stays that way.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
By default, we don't do any accounting. By specializing this class and providing
an accounter class, we can account how much memory are we reading as we read
through the elements.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
This is so we can template it without worrying about declaring the
specializations in the .cc file.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
The code that is common will live in its own reader, the iterator_reader. All
friendly private access to memtable attributes and methods happen through the
iterator reader.
After this patch, we are now left with the scanning_reader - same as always,
but now implemented on top of the iterator_reader, and a flush_reader, which
will be used by SSTable flushes only.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Right now the special reader doesn't do much, but the idea is that we will
soon replace it will a reader that specializes in flush, and is in turn able
to provide read-side on-flush functionality like virtual dirty.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We allocate objects of a certain size, but we use a bit more memory to hold
them. To get a clerer picture about how much memory will an object cost us, we
need help from the allocator. This patch exports an interface that allow users
to query into a specific allocator to get that information.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
In order to allow Scylla’s docker container to handle multiple network
interfaces, the start-scylla script was refactored:
- `$IP` is now called `$SCYLLA_LISTEN_ADDRESS`, so it is less likely to
be confused or interfere with other environment variables.
- `$SCYLLA_LISTEN_ADDRESS` now checks its value and also tries to
resolve a hostname, if no IP was set to it.
- `$SCYLLA_LISTEN_DEVICE` can now be set as environment variable and
contain any available NIC device name (e.g. `eth0`). The script
automatically retrieves the IP address from the device.
Usage:
1. With `$SCYLLA_LISTEN_ADDRESS` as IP:
`docker run -t -i --rm --name scylla -e SCYLLA_LISTEN_ADDRESS=192.168.1.100 scylladb/scylla`
2. With `$SCYLLA_LISTEN_ADDRESS` as hostname:
`docker run -t -i --rm --name scylla -e SCYLLA_LISTEN_ADDRESS=containername.network.lan scylladb/scylla`
3. With `$SCYLLA_LISTEN_DEVICE`:
`docker run -t -i --rm --name scylla -e SCYLLA_LISTEN_DEVICE=eth0 scylladb/scylla`
Message-Id: <20161003151230.67672-1-marius@twostairs.com>
Currently, the code responsible for calculating ranges for the next
request could produce a wrap-around partition range. For example, if the
original range was (unimportant, A] and the last partition key A then
the output range would be (A, A].
This patch adds checks to make sure that in such cases the range is
removed.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1475497244-2790-1-git-send-email-pdziepak@scylladb.com>
cassandra_exception::prepare_message() is called from derived classes'
constructors before the base cassnadra_exception object is constructed.
This is technically illegal but harmless. Fix by marking the function
static.
Found by clang.
"This patch set ensures we can correctly handle queries
where the minimum token is specified."
* 'min-token/v3' of github.com:duarten/scylla:
cql_query_test: Add test case for min/max token bounds
token_restriction: Deal with minimum tokens
partitioner: Parse token from bytes
This object, similarly to a global_schema_ptr, allows to dynamically
create the trace_state_ptr objects on different shards in a context
of the original tracing session.
This object would create a secondary tracing session object from the
original trace_state_ptr object when a trace_state_ptr object is needed
on a "remote" shard, similarly to what we do when we need it on a remote
Node.
Fixes#1678Fixes#1647
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Message-Id: <1474387767-21910-1-git-send-email-vladz@cloudius-systems.com>
This patch adds a call to the scylla-housekeeping check version during
setup, so a warning will be printed if a newer version is available.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This changes are for running scylla during setup.
It contains the following changes:
1. get the current version from the command line (as the syclla does not
run at this stage).
2. It support a mode parameter in the command line to indicate that we
running during the installation.
3. It accept an external uuid that will be used with all interaction
with the check_version server.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
When max sstable size is increased, higher levels are suffering from
starvation because we decide to compact a given level if the following
calculation results in a number greater than 1.001:
level_size(L) / max_size_for_level_l(L)
Fixes#1720.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Uniform token range distribution across sstables in a level > 1 was broken,
because we were only choosing sstable with lowest first key, when compacting
a level > 0. This resulted in performance problem because L1->L2 may have a
huge overlap over time, for example.
Last compacted key will now be stored for each level to ensure sort of
"round robin" selection of sstables for compactions at level >= 1.
That's also done by C*, and they were once affected by it as described in
https://issues.apache.org/jira/browse/CASSANDRA-6284.
Fixes#1719.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Paging code assumes that clustering row range [a, a] contains only one
row which may not be true. Another problem is that it tries to use
range<> interface for dealing with clustering key ranges which doesn't
work because of the lack of correct comparator.
Refs #1446.
Fixes#1684.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1475236805-16223-1-git-send-email-pdziepak@scylladb.com>
CQL server is supposed to throttle requests so that they don't
overflow memory. The problem is that it currently accounts for
request's memory only around reading of its frame from the connection
and not actual request execution. As a result too many requests may be
allowed to execute and we may run out of memory.
Fixes#1708.
Message-Id: <1475149302-11517-1-git-send-email-tgrabiec@scylladb.com>
This patch fixes a bug where queries such as the following are not
handled properly:
"SELECT * FROM ks.cf WHERE token(id) >
9207857967443869328 AND token(id) <= -9223372036854775808"
Here -9223372036854775808 represents the minimum token, which we were
just translating into a token with kind::key, thus returning incorrect
results.
Ref #1139
Ref #693Fixes#1717
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
This patch adds the from_bytes() function to the i_partitioner class,
whose purpose is parse a particular token and explicitly handle the
case when the minimum token is specified.
Signed-off-by: Duarte Nunes <duarte@scylladb.com>
Also add information about for how long has the oldest been sitting in the
queue. This is part of the backpressure work to allow us to throttle incoming
requests if we won't have memory to process them. Shortages can happen in all
sorts of places, and it is useful when designing and testing the solutions to
know where they are, and how bad they are.
This counter is named for consistency after similar counters from transport/.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Currently, we export the region group where memtables are placed as dirty bytes.
Upcoming patches will optimistically mark some bytes in this region as free, a
scheme we know as "virtual dirty".
We are still interested in knowing the real state of the dirty region, so we
will keep track of the bytes virtually freed and split the counters in two.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
2016-09-27 12:09:08 -04:00
546 changed files with 29317 additions and 8958 deletions
Use the [ScyllaDB user mailing list](https://groups.google.com/forum/#!forum/scylladb-users) for general questions and help.
# Reporting an issue
Please use the [Issue Tracker](https://github.com/scylladb/scylla/issues/) to report issues. Fill in as much information as you can in the issue template, especially for performance problems.
# Contributing Code to Scylla
To contribute code to Scylla, you need to sign the [Contributor License Agreement](http://www.scylladb.com/opensource/cla/) and send your changes as [patches](https://github.com/scylladb/scylla/wiki/Formatting-and-sending-patches) to the [mailing list](https://groups.google.com/forum/#!forum/scylladb-dev). We don't accept pull requests on GitHub.
throwexceptions::invalid_request_exception(sprint("Cannot rename unknown column %s in table %s",from,schema.cf_name()));
}
if(schema.get_column_definition(to.name())){
throwexceptions::invalid_request_exception(sprint("Cannot rename column %s to %s in table %s; another column of that name already exist",from,to,schema.cf_name()));
}
if(def->is_part_of_cell_name()){
throwexceptions::invalid_request_exception(sprint("Cannot rename non PRIMARY KEY part %s",from));
}
if(def->is_indexed()){
throwexceptions::invalid_request_exception(sprint("Cannot rename column %s because it is secondary indexed",from));
throwexceptions::invalid_request_exception(sprint("Cannot rename unknown column %s in table %s",from,column_family()));
}
if(schema->get_column_definition(to->name())){
throwexceptions::invalid_request_exception(sprint("Cannot rename column %s to %s in table %s; another column of that name already exist",from,to,column_family()));
}
if(def->is_part_of_cell_name()){
throwexceptions::invalid_request_exception(sprint("Cannot rename non PRIMARY KEY part %s",from));
}
if(def->is_indexed()){
throwexceptions::invalid_request_exception(sprint("Cannot rename column %s because it is secondary indexed",from));
}
validate_column_rename(*schema,*from,*to);
cfm.with_column_rename(from->name(),to->name());
// If the view includes a renamed column, it must be renamed in the view table and the definition.
throwexceptions::invalid_request_exception(sprint("Cannot add new field %s to type %s: a field of the same name already exists",_field_name->name(),_name.to_string()));
throwexceptions::invalid_request_exception(sprint("Type %s in incompatible with previous type %s of field %s in user type %s",_field_type->to_string(),previous->as_cql3_type()->to_string(),_field_name->name(),_name.to_string()));
throwexceptions::configuration_exception(sstring("Missing sub-option '")+compression_parameters::SSTABLE_COMPRESSION+"' for the '"+KW_COMPRESSION+"' option.");
throwexceptions::invalid_request_exception("COMPACT STORAGE with non-composite PRIMARY KEY require one column not part of the PRIMARY KEY, none given");
throwexceptions::invalid_request_exception("Only clustering key columns can be defined in CLUSTERING ORDER directive");
}
inti=0;
for(auto&pair:_defined_ordering){
for(auto&pair:_properties.defined_ordering()){
auto&id=pair.first;
auto&c=_column_aliases.at(i);
if(!(*id==*c)){
if(find_ordering_info(c)){
if(_properties.find_ordering_info(c)){
throwexceptions::invalid_request_exception(sprint("The order of columns in the CLUSTERING ORDER directive must be the one of the clustering key (%s must appear before %s)",c,id));
}else{
throwexceptions::invalid_request_exception(sprint("Missing CLUSTERING ORDER for column %s",c));
throwexceptions::invalid_request_exception(sprint("Cannot drop user type %s.%s as it is still used by user type %s",keyspace,type->get_name_as_string(),ut->get_name_as_string()));
throwexceptions::invalid_request_exception(sprint("Cannot drop user type %s.%s as it is still used by table %s.%s",keyspace,type->get_name_as_string(),cfm->ks_name(),cfm->cf_name()));
}
}
}
}catch(no_such_keyspace&e){
throwexceptions::invalid_request_exception(sprint("Cannot drop type in unknown keyspace %s",keyspace()));
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
