This series adds background reclaim to lsa, with the goal
that most large allocations can be satisfied from available
free memory, and and reclaim work can be done from a preemptible
context.
If the workload has free cpu, then background reclaim will
utilize that free cpu, reducing latency for the main workload.
Otherwise, background reclaim will compete with the main
workload, but since that work needs to happen anyway,
throughput will not be reduced.
A unit test is added to verify it works.
Fixes#1634.
Closes#8044
* github.com:scylladb/scylla:
test: logalloc_test: test background reclaim
logalloc: reduce gap between std min_free and logalloc min_free
logalloc: background reclaim
logalloc: preemptible reclaim
Test that the background reclaimer is able to compete with a
fake load and reclaim 10 MB/s. The test is quite stressful as the "LRU"
is fully randomized.
If the background reclaimer is disabled, the test fails as soon as the
20MB "gap" is exhausted. With the reclaimer enabled, it is able to
free memory ahead of the allocations.
"
Current storage of cells in a row is a union of vector and set. The
vector holds 5 cell_and_hash's inline, up to 32 ones in the external
storage and then it's switched to std::set. Once switched, the whole
union becomes the waste of space, as it's size is
sizeof(vector head) + 5 * sizeof(cell and hash) = 90+ bytes
and only 3 pointers from it are used (std::set header). Also the
overhead to keep cell_and_hash as a set entry is more then the size
of the structure itself.
Column ids are 32-bit integers that most likely come sequentialy.
For this kind of a search key a radix tree (with some care for
non-sequential cases) can be beneficial.
This set introduces a compact radix tree, that uses 7-bit sub values
from the search key to index on each node and compacts the nodes
themselves for better memory usage. Then the row::_storage is replaced
with the new tree.
The most notable result is the memory footprint decrease, for wide
rows down to 2x times. The performance of micro-benchmarks is a bit
lower for small rows and (!) higer for longer (8+ cells). The numbers
are in patch #12 (spoiler: they are better than for v2)
v3:
- trimmed size of radix down to 7 bits
- simplified the nodes layouts, now there are 2 of them (was 4)
- enhanced perf_mutation to test N-cells schema
- added AVX intra-nodes search for medium-sized nodes
- added .clone_from() method that helped to improve perf_mutation
- minor
- changed functions not to return values via refs-arguments
- fixed nested classes to properly use language constructors
- renamed index_to to key_t to distinguish from node_index_t
- improved recurring variadic templates not to use sentinel argument
- use standard concepts
v2:
- fixed potential mis-compilation due to strict-aliasing violation
- added oracle test (radix tree is compared with std::map)
- added radix to perf_collection
- cosmetic changes (concepts, comments, names)
A note on item 1 from v2 changelog. The nodes are no longer packed
perfectly, each has grown 3 bytes. But it turned out that when used
as cells container most of this growth drowned in lsa alignments.
next todo:
- aarch64 version of 16-keys node search
tests: unit(dev), unit(debug for radix*), pref(dev)
"
* 'br-radix-tree-for-cells-3' of https://github.com/xemul/scylla:
test/memory_footpring: Print radix tree node sizes
row: Remove old storages
row: Prepare row::equal for switch
row: Prepare row::difference for switch
row: Introduce radix tree storage type
row-equal: Re-declare the cells_equal lambda
test: Add tests for radix tree
utils: Compact radix tree
array-search: Add helpers to search for a byte in array
test/perf_collection: Add callback to check the speed of clone
test/perf_mutation: Add option to run with more than 1 columns
test/perf_mutation: Prepare to have several regular columns
test/perf_mutation: Use builder to build schema
Until now, the lists of streams in the `cdc_streams_descriptions` table
for a given generation were stored in a single collection. This solution
has multiple problems when dealing with large clusters (which produce
large lists of streams):
1. large allocations
2. reactor stalls
3. mutations too large to even fit in commitlog segments
This commit changes the schema of the table as described in issue #7993.
The streams are grouped according to token ranges, each token range
being represented by a separate clustering row. Rows are inserted in
reasonably large batches for efficiency.
The table is renamed to enable easy upgrade. On upgrade, the latest CDC
generation's list of streams will be (re-)inserted into the new table.
Yet another table is added: one that contains only the generation
timestamps clustered in a single partition. This makes it easy for CDC
clients to learn about new generations. It also enables an elegant
two-phase insertion procedure of the generation description: first we
insert the streams; only after ensuring that a quorum of replicas
contains them, we insert the timestamp. Thus, if any client observes a
timestamp in the timestamps table (even using a ONE query),
it means that a quorum of replicas must contain the list of streams.
---
Nodes automatically ensure that the latest CDC generation's list of
streams is present in the streams description table. When a new
generation appears, we only need to update the table for this
generation; old generations are already inserted.
However, we've changed the description table (from
`cdc_streams_descriptions` to `cdc_streams_descriptions_v2`). The
existing mechanism only ensures that the latest generation appears in
the new description table. We add an additional procedure that
rewrites the older generations as well, if we find that it is necessary
to do so (i.e. when some CDC log tables may contain data in these
generations).
Closes#8116
* github.com:scylladb/scylla:
tests: add a simple CDC cql pytest
cdc: add config option to disable streams rewriting
cdc: rewrite streams to the new description table
cql3: query_processor: improve internal paged query API
cdc: introduce no_generation_data_exception exception type
docs: cdc: mention system.cdc_local table
cdc: coroutinize do_update_streams_description
sys_dist_ks: split CDC streams table partitions into clustered rows
cdc: use chunked_vector for streams in streams_version
cdc: remove `streams_version::expired` field
system_distributed_keyspace: use mutation API to insert CDC streams
storage_service: don't use `sys_dist_ks` before it is started
The `query_processor::query` method allowed internal paged queries.
However, it was quite limited, hardcoding a number of parameters:
consistency level, timeout config, page size.
This commit does the following improvements:
1. Rename `query` to `query_internal` to make it obvious that this API
is supposed to be used for internal queries only
2. Extend the method to take consistency level, timeout config, and page
size as parameters
3. Remove unused overloads of `query_internal`
4. Fix a bunch of typos / grammar issues in the docstring
Until now, the lists of streams in the `cdc_streams_descriptions` table
for a given generation were stored in a single collection. This solution
has multiple problems when dealing with large clusters (which produce
large lists of streams):
1. large allocations
2. reactor stalls
3. mutations too large to even fit in commitlog segments
This commit changes the schema of the table as described in issue #7993.
The streams are grouped according to token ranges, each token range
being represented by a separate clustering row. Rows are inserted in
reasonably large batches for efficiency.
The table is renamed to enable easy upgrade. On upgrade, the latest CDC
generation's list of streams will be (re-)inserted into the new table.
Yet another table is added: one that contains only the generation
timestamps clustered in a single partition. This makes it easy for CDC
clients to learn about new generations. It also enables an elegant
two-phase insertion procedure of the generation description: first we
insert the streams; only after ensuring that a quorum of replicas
contains them, we insert the timestamp. Thus, if any client observes a
timestamp in the timestamps table (even using a ONE query),
it means that a quorum of replicas must contain the list of streams.
Test log consistency after apply_snapshot() is called.
Ensure log::last_term() log::last_conf_index() and log::size()
work as expected.
Misc cleanups.
* scylla-dev/raft-confchange-test:
raft: add a unit test for voting
raft: do not account for the same vote twice
raft: remove fsm::set_configuration()
raft: consistently use configuration from the log
raft: add ostream serialization for enum vote_result
raft: advance commit index right after leaving joint configuration
raft: add tracker test
raft: tidy up follower_progress API
raft: update raft::log::apply_snapshot() assert
raft: add a unit test for raft::log
raft: rename log::non_snapshoted_length() to log::length()
raft: inline raft::log::truncate_tail()
raft: ignore AppendEntries RPC with a very old term
raft: remove log::start_idx()
raft: return a correct last term on an empty log
raft: do not use raft::log::start_idx() outside raft::log()
raft: rename progress.hh to tracker.hh
raft: extend single_node_is_quiet test
`_range_override` is used to store the modified range the reader reads
after it has to be recreated (when recreating a reader it's read range
is reduced to account for partitions it already read). When engaged,
this field overrides the `_pr` field as the definitive range the reader
is supposed to be currently reading. Fast forwarding conceptually
overrides the range the reader is currently reading, however currently
it doesn't reset the `_range_override` field. This resulted in
`_range_override` (containing the modified pre-fast-forward range)
incorrectly overriding the fast-forwarded-to range in `_pr` when
validating the first partition produced by the just recreated reader,
resulting in a false-positive validation failure.
Fixes: #8059
Tests: unit(release)
Signed-off-by: Botond Dénes <bdenes@scylladb.com>
Message-Id: <20210217164744.420100-1-bdenes@scylladb.com>
Unlike flat_mutation_reader_opt that is defined using
optimized_optional<flat_mutation_reader>, std::optional<T> does not evaluate
to `false` after being moved, only after it is explicitly reset.
Use flat_mutation_reader_opt rather than std::optional<flat_mutation_reader>
to make it easier to check if it was closed before it's destroyed
or being assigned-over.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20210215101254.480228-6-bhalevy@scylladb.com>
Currently, whole topology description for CDC is stored in a single row.
This means that for a large cluster of strong machines (say 100 nodes 64
cpus each), the size of the topology description can reach 32MB.
This causes multiple problems. First of all, there's a hard limit on
mutation size that can be written to Scylla. It's related to commit log
block size which is 16MB by default. Mutations bigger than that can't be
saved. Moreover, such big partitions/rows cause reactor stalls and
negatively influence latency of other requests.
This patch limits the size of topology description to about 4MB. This is
done by reducing the number of CDC streams per vnode and can lead to CDC
data not being fully colocated with Base Table data on shards. It can
impact performance and consistency of data.
This is just a quick fix to make it easily backportable. A full solution
to the problem is under development.
For more details see #7961, #7993 and #7985.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Closes#8048
* github.com:scylladb/scylla:
cdc: Limit size of topology description
cdc: Extract create_stream_ids from topology_description_generator
Currently, whole topology description for CDC is stored in a single row.
This means that for a large cluster of strong machines (say 100 nodes 64
cpus each), the size of the topology description can reach 32MB.
This causes multiple problems. First of all, there's a hard limit on
mutation size that can be written to Scylla. It's related to commit log
block size which is 16MB by default. Mutations bigger than that can't be
saved. Moreover, such big partitions/rows cause reactor stalls and
negatively influence latency of other requests.
This patch limits the size of topology description to about 4MB. This is
done by reducing the number of CDC streams per vnode and can lead to CDC
data not being fully colocated with Base Table data on shards. It can
impact performance and consistency of data.
This is just a quick fix to make it easily backportable. A full solution
to the problem is under development.
For more details see #7961, #7993 and #7985.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Commit aab6b0ee27 introduced the
controversial new IMR format, which relied on a very template-heavy
infrastructure to generate serialization and deserialization code via
template meta-programming. The promise was that this new format, beyond
solving the problems the previous open-coded representation had (working
on linearized buffers), will speed up migrating other components to this
IMR format, as the IMR infrastructure reduces code bloat, makes the code
more readable via declarative type descriptions as well as safer.
However, the results were almost the opposite. The template
meta-programming used by the IMR infrastructure proved very hard to
understand. Developers don't want to read or modify it. Maintainers
don't want to see it being used anywhere else. In short, nobody wants to
touch it.
This commit does a conceptual revert of
aab6b0ee27. A verbatim revert is not
possible because related code evolved a lot since the merge. Also, going
back to the previous code would mean we regress as we'd revert the move
to fragmented buffers. So this revert is only conceptual, it changes the
underlying infrastructure back to the previous open-coded one, but keeps
the fragmented buffers, as well as the interface of the related
components (to the extent possible).
Fixes: #5578
The off-by-one error would cause
test_multishard_combining_reader_non_strictly_monotonic_positions to fail if
the added range_tombstones filled the buffer exactly to the end.
In such situation, with the old loop condition,
make_fragments_with_non_monotonic_positions would add one range_tombstone too
many to the deque, violating the test assumptions.
Due to small value optimizations, the removed assertions are not true in
general. Until now, atomic_cell did not use small value optimizations, but
it will after upcoming changes.
sstable_run_based_compaction_test assumed that sstables are freed immediately
after they are fully processed.
Hovewer, since commit b524f96a74,
mutation_reader_merger releases sstables in batches of 4, which breaks the
assumption. This fix adjusts the test accordingly.
Until now, the test only kept working by chance: by coincidence, the number of
test sstables processed by merging_reader in a single fill_buffer() call was
divisible by 4. Since the test checks happen between those calls,
the test never witnessed a situation when an sstable was fully processed,
but not released yet.
The error was noticed during the work on an upcoming patch which changes the
size of mutation_fragment, and reduces the number of test sstables processed
in a single fill_buffer() call, which breaks the test.
raft::log::start_idx() is currently not meaningful
in case the log is empty.
Avoid using it in fsm::replicate_to() and avoid manual search for
previous log term, instead encapsulate the search in log::term_for().
As a side effect we currently return a correct term (0)
when log matching rule is exercised for an empty log
and the very first snapshot with term 0. Update raft_etcd_test.cc
accordingly.
This change happens to reduce the overall line count.
While at it, improve the comments in raft::replicate_to().
"
Currently, register_inactive_read accepts an eviction_notify_handler
to be called when the inactive_read is evicted.
However, in case there was an error in register_inactive_read
the notification function isn't called leaving behind
state that needs to be cleaned up.
This series separates the register_inactive_reader interface
into 2 parts:
1. register_inactive_reader(flat_mutation_reader) - which just registers
the reader and return an inactive_read_handle, *if permitted*.
Otherwise, the notification handler is not called (it is not known yet)
and the caller is not expected to do anything fance at this point
that will require cleanup.
This optimizes the server when overloaded since we do less work
that we'd need to undo in case the reader_concurrecy_semaphore
runs out of resources.
2. After register_inactive_reader succeeded to return a valid
inactive_read_handle, the caller sets up its local state
and may call `set_notify_handler` to set the optional
notify_handler and ttl on the o_r_h.
After this state, the notify_handler will be called when
the inactive_reader is evicted, for any reason.
querier_cache::insert_querier was modified to use the
above procedure and to handle (and log/ignore) any error
in the process.
inactive_read_handle and inactive_read keeping track of each other
was simplified by keeping an iterator in the handle and a backpointer
in the inactive_read object. The former is used to evict the reader
and to set the notify_handler and/or ttl without having to lookup the i_r.
The latter is used to invalidate the i_r_h when the i_r is destroyed.
Test: unit(release), querier_cache_test(debug)
"
* tag 'register_inactive_read-error-handling-v6' of github.com:bhalevy/scylla:
querier_cache: insert_querier: ignore errors to register inactive reader
querier_cache: insert_querier: handle errors
querier_utils: mark functions noexcept
reader_concurrency_semaphore: register_inactive_read: make noexcept
reader_concurrency_semaphore: separate set_notify_handler from register_inactive_reader
reader_concurrency_semaphore: inactive_read: make ttl_timer non-optional
reader_concurrency_semaphore: inactive_read: use intrusive list
reader_concurrency_semaphore: do_wait_admission: use try_evict_one_inactive_read
reader_concurrency_semaphore: try_evict_one_inactive_read: pass evict_reason
reader_concurrency_semaphore: unregister_inactive_read: calling on wrong semaphore is an internal error
reader_concurrency_semaphore: unregister_inactive_read: do nothing if disengaged
reader_concurrency_semaphore: inactive_read_handle: swap definition order
reader_lifecycle_policy: retire low level try_resume method
reader_concurrency_semaphore: inactive_read: keep a flat_mutation_reader
The same trick is used as in C*:
79e693e16e/src/java/org/apache/cassandra/locator/NetworkTopologyStrategy.java (L241)
The edited CQL test relied on quietly accepting non-existing DCs, so it had to
be removed. Also, one boost-test referred to nonexistent `datacenter2` and had
to be removed.
Fixes#7595Closes#8056
* github.com:scylladb/scylla:
tests: Adjusted tests for DC checking in NTS
locator: Check DC names in NTS
CQL test relied on quietly acceptiong non-existing DCs, so it had
to be removed. Also, one boost-test referred to nonexisting
`datacenter2` and had to be removed.
Register the inactive reader first with no
evict_notify_handler and ttl.
Those can be set later, only if registration succeeded.
Otherwise, as in the querier example, there is no need
to to place the querier in the index and erase it
on eviction.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Calling unregister_inactive_read on the wrong semaphore is a blatant
bug so better call on_internal_error so it'd be easier to catch and fix.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Fix raft_fsm_test failure in debug mode. ASAN complained
that follower_progress is used in append_entries_reply()
after it was destroyed. This could happen if in maybe_commit()
we switched to a new configuration and destroyed old progress
objects.
The fix is to lookup the object one more time after maybe_commit().
commitlog was changed to use fragmented_temporary_buffer::ostream (db::commitlog::output).
So if there are discontiguous small memory blocks, they can be used to satisfy
an allocation even if no contiguous memory blocks are available.
To prevent that, as Avi suggested, this change allocates in 128K blocks
and frees the last one to succeed (so that we won't fail on allocating continuations).
Fixes#8028
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20210203100333.862036-1-bhalevy@scylladb.com>
"
Currently inactive readers are stored in two different places:
* reader concurrency semaphore
* querier cache
With the latter registering its inactive readers with the former. This
is an unnecessarily complex (and possibly surprising) setup that we want
to move away from. This series solves this by moving the responsibility
if storing of inactive reads solely to the reader concurrency semaphore,
including all supported eviction policies. The querier cache is now only
responsible for indexing queriers and maintaining relevant stats.
This makes the ownership of the inactive readers much more clear,
hopefully making Benny's work on introducing close() and abort() a
little bit easier.
Tests: unit(release, debug:v1)
"
* 'unify-inactive-readers/v2' of https://github.com/denesb/scylla:
reader_concurrency_semaphore: store inactive readers directly
querier_cache: store readers in the reader concurrency semaphore directly
querier_cache: retire memory based cache eviction
querier_cache: delegate expiry to the reader_concurrency_semaphore
reader_concurrency_semaphore: introduce ttl for inactive reads
querier_cache: use new eviction notify mechanism to maintain stats
reader_concurrency_semaphore: add eviction notification facility
reader_concurrency_semaphore: extract evict code into method evict()
This is the continuaiton of the row-cache performance
improvements, this time -- the rework of clustering keys part.
The goal is to solve the same set of problems:
- logN eviction complexity
- deep and sparse tree
Unlike partitions, this cache has one big feature that makes it
impossible to just use existing B+ tree:
There's no copyable key at hands. The clustering key is the
managed_bytes() that is not nothrow-copy-constructibe, neither
it's hash-able for lookup due to prefix lookup.
Thus the choice is the B-tree, which is also N-ary one, but
doesn't copy keys around.
B-trees are like B+, but can have key:data pairs in inner nodes,
thus those nodes may be significantly bigger then B+ ones, that
have data-s only in leaf trees. Not to make the memory footprint
worse, the tree assumes that keys and data live on the same object
(the rows_entry one), and the tree itself manages only the key
pointers.
Not to invalidate iterators on insert/remove the tree nodes keep
pointers on keys, not the keys themselves.
The tree uses tri-compare instead of less-compare. This makes the
.find and .lower_bound methods do ~10% less comparisons on random
insert/lookup test.
Numbers:
- memory_footprint: B-tree master
rows_entry size: 216 232
1 row
in-cache: 968 960 (because of dummy entry)
in-memtable: 1006 1022
100 rows
in-cache: 50774 50856
in-memtable: 50620 50918
- mutation_test: B-tree master
tps.average: 891177 833896
- simple_query: B-tree master
tps.median: 71807 71656
tps.maximum: 71847 71708
* xemul/clustering-cache-over-btree-4:
mutation_partition: Save one keys comparison
partition_snapshot_row_cursor: Remove rows pointer
mutation_partition: Use B-tree insertion sugar
perf-test : Print B-tree sizes
mutation_partition: Switch cache of rows onto B-tree
partition_snapshot_reader: Rename cmp to less for explicity
mutation_partition: Make insertion bullet-proof
mutation_partition: Use tri-compare in non-set places
flat_mutation_reader: Use clear() in destroy_current_mutation()
rows_entry: Generalize compare
utils: Intrusive B-tree (with tests)
tests: Generalize bptree compaction test
tests: Generalize bptree stress test
Fixes#7615
Makes the CL writer interface N-valued (though still 1 for the "old" paths). Adds a new write path to input N mutations -> N rp_handles.
Guarantees that all entries are written or none are, and that they will be flushed to disk together.
Small test included.
Closes#7616
* github.com:scylladb/scylla:
commitlog_test: Add multi-entry write test
commitlog: Add "add_entries" call to allow inputting N mutations
commitlog: Make commitlog entries optionally multi-entry
commitlog: Move entry_writer definition to cc file
Whenever we push a fragment, we check whether the buffer is
full and return proceed::no if so, so that the state machine pauses
and lets the consumer continue. This patch adds an additional
condition - if preemption is needed, we also return proceed::no.
This drops us back to the outer loop
(in sstable_mutation_reader::fill_buffer), which will yield to
the reactor as part of seastar::do_until().
Two cases (partition_start and partition_end) did not have the
check for is_buffer_full(); it is added now. This can trigger
is the partition has no rows.
Unlike the previous attempt, push_ready_fragments() is not touched.
The extra preemption opportunities triggered a preexisting bug in
clustering_ranges_walker; it is fixed in the first patch of the series.
I tested this by reading from a large partition with a simple
schema (pk int, ck int, primary key(pk, ck)) with BYPASS CACHE.
However, even without the patch I only got sporadic stalls
with the detector set to 1ms, so it's possible I'm not testing
correctly.
Test: unit (dev, debug, release)
Fixes#7883.
Closes#7928
* github.com:scylladb/scylla:
sstable: reader: preempt after every fragment
clustering_range_walker: fix false discontiguity detected after a static row
Rather than asserting, as seen in #7977.
This shouldn't crash the server in production.
Add unit test that reproduces this scenario
and verifies the internal error exception.
Fixes#7977
Test: unit(release)
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20210201163051.1775536-1-bhalevy@scylladb.com>
"
This series extends the scylla_metadata sstable component
to hold an optional testual description of the sstable origin.
It describes where the sstables originated from
(e.g. memtable, repair, streaming, compaction, etc.)
The origin string is provided by the sstable writer via
sstable_writer_config, written to the scylla_metadata component,
and loaded on sstable::load().
A get_origin() method was added to class sstable to retrieve
its origin. It returns an empty string by default if the origin
is missing.
Compaction now logs the sstable origin for each sstable it
compacts, and it generates the sstable origin for all sstables
in generates. Regular compaction origin is simply set to "compaction"
while other compaction types are mentioned by name, as
"cleanup", "resharding", "reshaping", etc.
A unit test was added to test the sstable_origin by writing either
an empty origin and a random string, and then comparing
the origin retrieved by sstable::load to the one written.
Test: unit(release)
Fixes#7880
"
* tag 'sstable-origin-v2' of github.com:bhalevy/scylla:
compaction: log sstable origin
sstables: scylla_metadata: add support for sstable_origin
sstables: sstable_writer_config: add origin member
The switch is pretty straightforward, and consists of
- change less-compare into tri-compare
- rename insert/insert_check into insert_before_hint
- use tree::key_grabber in mutation_partition::apply_monotonically to
exception-safely transfer a row from one tree to another
- explicitly erase the row from tree in rows_entry::on_evicted, there's
a O(1) tree::iterator method for this
- rewrite rows_entry -> cache_entry transofrmation in the on_evicted to
fit the B-tree API
- include the B-tree's external memory usage into stats
That's it. The number of keys per node was is set to 12 with linear search
and linear extention of 20 because
- experimenting with tree shows that numbers 8 through 10 keys with linear
search show the best performance on stress tests for insert/find-s of
keys that are memcmp-able arrays of bytes (which is an approximation of
current clustring key compare). More keys work slower, but still better
than any bigger value with any type of search up to 64 keys per node
- having 12 keys per nodes is the threshold at which the memory footprint
for B-tree becomes smaller than for boost::intrusive::set for partitions
with 32+ keys
- 20 keys for linear root eats the first-split peak and still performs
well in linear search
As a result the footpring for B tree is bigger than the one for BST only for
trees filled with 21...32 keys by 0.1...0.7 bytes per key.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
The design of the tree goes from the row-cache needs, which are
1. Insert/Remove do not invalidate iterators
2. Elements are LSA-manageable
3. Low key overhead
4. External tri-comparator
5. As little actions on insert/remove as possible
With the above the design is
Two types of nodes -- inner and leaf. Both types keep pointer on parent nodes
and N pointers on keys (not keys themselves). Two differences: inner nodes have
array of pointers on kids, leaf nodes keep pointer on the tree (to update left-
and rightmost tree pointers on node move).
Nodes do not keep pointers/references on trees, thus we have O(1) move of any
object, but O(logN) to get the tree size. Fortunately, with big keys-per-node
value this won't result in too many steps.
In turn, the tree has 3 pointers -- root, left- and rightmost leaves. The latter
is for constant-time begin() and end().
Keys are managed by user with the help of embeddable member_hook instance,
which is 1 pointer in size.
The code was copied from the B+ tree one, then heavily reworked, the internal
algorythms turned out to differ quite significantly.
For the sake of mutation_partition::apply_monotonically(), which needs to move
an element from one tree into another, there's a key_grabber helping wrapper
that allows doing this move respecting the exception-safety requirement.
As measured by the perf_collections test the B-tree with 8 keys is faster, than
the std::set, but slower than the B+tree:
vs set vs b+tree
fill: +13% -6%
find: +23% -35%
Another neat thing is that 1-key insertion-removal is ~40% faster than
for BST (the same number of allocations, but the key object is smaller,
less pointers to set-up and less instructions to execute when linking
node with root).
v4:
- equip insertion methods with on_alloc_point() calls to catch
potential exception guarantees violations eariler
- add unlink_leftmost_without_rebalance. The method is borrowed from
boost intrusive set, and is added to kill two birds -- provide it,
as it turns out to be popular, and use a bit faster step-by-step
tree destruction than plain begin+erase loop
v3:
- introduce "inline" root node that is embedded into tree object and in
which the 1st key is inserted. This greatly improves the 1-key-tree
performance, which is pretty common case for rows cache
v2:
- introduce "linear" root leaf that grows on demand
This improves the memory consumption for small trees. This linear node may
and should over-grow the NodeSize parameter. This comes from the fact that
there are two big per-key memory spikes on small trees -- 1-key root leaf
and the first split, when the tree becomes 1-key root with two half-filled
leaves. If the linear extention goes above NodeSize it can flatten even the
2nd peak
- mitigate the keys indirection a bit
Prefetching the keys while doing the intra-node linear scan and the nodes
while descending the tree gives ~+5% of fill and find
- generalize stress tests for B and B+ trees
- cosmetic changes
TODO:
- fix few inefficincies in the core code (walks the sub-tree twice sometimes)
- try to optimize the leaf nodes, that are not lef-/righmost not to carry
unused tree pointer on board
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
clustering_range_walker detects when we jump from one row range to another. When
a static row is included in the query, the constructor sets up the first before/after
bounds to be exactly that static row. That creates an artificial range crossing if
the first clustering range is contiguous with the static row.
This can cause the index to be consulted needlessly if we happen to fall back
to sstable_mutation_reader after reading the static row.
A unit test is added.
Ref #7883.
Support configuration changes based on joint consensus.
When a user adds a configuration entry, commit an interim "joint
consensus" configuration to the log first, and transition to the
final configuration once both C_old and C_new configurations
accept the joint entry.
Misc cleanups.
* scylla-dev/raft-config-changes-v2:
raft: update README.md
raft: add a simple test for configuration changes
raft: joint consensus, wire up configuration changes in the API
raft: joint consensus, count votes using joint config
raft: joint consensus, wire up configuration changes in FSM
raft: joint consensus, update progress tracker with joint configuration
raft: joint consensus, don't store configuration in FSM
raft: joint consensus, keep track of the last confchange index in the log
raft: joint consensus, implement helpers in class configuration
raft: joint consensus, use unordered_set for server_address list
raft: joint consensus, switch configuration to joint
raft: rename check_committed() to maybe_commit()
raft: fix spelling and add comments
Add new scylla_metadata_type::SSTableOrigin.
Store and retrive a sstring to the scylla metadata component.
Pass sstable_writer_config::origin from the mx sstable writer
and ignore it in the k_l writer.
Add unit test to verify the sstable_origin extension
using both empty and a random string.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Add a string describing where the sstables originated
from (e.g. memtable, repair, streaming, compaction, etc.)
If configure_writer is called with a nullptr, the origin
will be equal to an empty string.
Introduce test_env_sstables_manager that provides an overload
of configure_writer with no parmeters that calls the base-class'
configure_writer with "test" origin. This was to reduce the
code churn in this patch and to keep the tests simple.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
The idea of the monotonicity checking test is: try to apply
one one random partition to another random one sequentually
failing allocations. Each time allocation fails (with the
bad_alloc exception) -- check the exception guarantee is
respected, then apply (!) the very same two partitions to
each other. At the end of the test we make sure, that an
exception may pop up at any point of application and it
will be safe.
This idea is flawed currently. When verifying the guarantee
the test moves the 2nd partition and leaves it empty for the
next loop iteration. So right on the 2nd attempt to apply
partitions it becomes a no-op, doesn't fail and no more
exceptions arise.
Fix by restoring both partitions at the end of each check.
Broken since 74db08165d.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Message-Id: <20210129153641.5449-1-xemul@scylladb.com>
This series contains an initial implementation of raft persistency module
that uses `raft` system table as the underlying storage model.
"system.raft" table will be used as a backend storage for implementing
raft persistence module in Scylla. It combines both raft log,
persisted vote and term, and snapshot info.
The table is partitioned by group id, thus allowing multi-raft
operation. The rest of the table structure mirrors the fields of
corresponding core raft structures defined in `raft.hh`, such as
`raft::log_entry`.
The raft table stores the only the latest snapshot id while
the actual snapshot will be available in a separate table
called `system.raft_snapshots`. The schema of `raft_snapshots`
mirrors the fields of `raft::snapshot` structure.
IDL definitions are also added for every raft struct so that we
automatically provide serialization and deserialization facilities
needed both for persistency module and for future RPC implmementation.
The first patch is a side-change needed to provide complete
serialization/deserialization for `bytes_ostream`, which we
need when persisting the raft log in the table (since `data`
is a variant containing `raft::command` (aka `bytes_ostream`)
among others).
`bytes_ostream` was lacking `deserialize` function, which is
added in the patch.
The second patch provides serializer for `lw_shared_ptr<T>`
which will be used for `raft::append_entries`, which has
a field with `std::vector<const lw_shared_ptr<raft::log_entry>>`
type.
There is also a patch to extend `fragmented_temporary_buffer`
with a static function `allocate_to_fit` that allocates an
instance of the fragmented buffer that has a specified size.
Individual fragment size is limited to 128kb.
The patch-set also contains the test suite covering basic
functionality of the persistency module.
* manmanson/raft-api-impl-v11:
raft/sys_table_storage: add basic tests for raft_sys_table_storage
raft: introduce `raft_sys_table_storage` class
utils: add `fragmented_temporary_buffer::allocate_to_fit`
raft: add IDL definitions for raft types
raft: create `system.raft` and `system.raft_snapshots` tables
serializer: add `serializer<lw_shared_ptr<T>>` specialization
serializer: add `deserialize` function overload for `bytes_ostream`
The test suite covers the most basic use cases for the system table
backed raft persistency module:
* store/load vote and term
* store/load snapshot
* store snapshot with log tail truncation
* store/load log entries
* log truncation
Tests: unit(dev)
Signed-off-by: Pavel Solodovnikov <pa.solodovnikov@scylladb.com>
