This will make migration to flat_mutation_reader much
easier and sstables::mutation_reader is going away with
this migration anyway.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
Restrict readers based on their memory consumption, instead of the count
of the top-level readers. To do this an interposer is installed at the
input_stream level which tracks buffers emmited by the stream. This way
we can have an accurate picture of the readers' actual memory
consumption.
New readers will consume 16k units from the semaphore up-front. This is
to account their own memory-consumption, apart from the buffers they
will allocate. Creating the reader will be deferred to when there are
enough resources to create it. As before only new readers will be
blocked on an exhausted semaphore, existing readers can continue to
work.
Restrict readers based on their memory consumption, instead of the count
of the top-level readers. To do this an interposer is installed at the
input_stream level which tracks buffers emmited by the stream. This way
we can have an accurate picture of the readers' actual memory
consumption.
New readers will consume 16k units from the semaphore up-front. This is
to account their own memory-consumption, apart from the buffers they
will allocate. Creating the reader will be deferred to when there are
enough resources to create it. As before only new readers will be
blocked on an exhausted semaphore, existing readers can continue to
work.
To reduce the memory footprint of compression-info, n offsets are
grouped together into segments, where each segment stores a base
absolute offset into the file, the other offsets in the segments being
relative offsets (and thus of reduced size). Also offsets are
allocated only just enough bits to store their maximum value. The
offsets are thus packed in a buffer like so:
arrrarrrarrr...
where n is 4, a is an absolute offset and r are offsets relative to a.
The optimal value of n can be calculated for a given file_size (f) and
chunk_size (c), by finding the minima of the following function:
f(n) = (f/c)/n * (log2(f) + (n - 1)*log2((n-1)*(c + 64)))
This is done in an empirical way, using a script (see below).
Furthermore segments are stored in buckets, where each bucket has its
own base offset. Each bucket therefore can address an equal chunk of the
file and furthermore each segment in a bucket can address an equal
sub-chunk of this area.
The value of a given offset i is thus:
bucket_base_offset_for(i) + segment_base_offset_for(i) + offset(i)
To account for the bucketed storage we calculate a local_f, which is
optimized so that a bucketful of segmented offsets can address the
largest possible chunk of f. As value of this local_f only depends on
the bucket_size (b) and c the value of n can be made independent of f
and therefore only depend on one dynamic value, c. This makes life much
simpler as we don't need to know the size of the file up-front, we can
just append buckets to the storage on demand, while the required storage
is still less than a third [1] of the original storage requirements
(std::deque<uint64>).
The table with the minima(f(n)) for different f and c values is
pre-computed by gen_segmented_compress_params.py and
stored in sstables/segmented_compress_params.hh. This script also
creates a table with the best values of local_f for the given
bucket_size. At runtime we only select the best params based on c.
[1] This was calculated for c=4K and b=4K
Currently, a summary entry is added after min_index_interval index
entries were written. Not taking into account size of index entries
becomes a problem with large partitions which may create big index
entries due to promoted indexes. Read performance is affected as a
consequence because index entries spanned by summary are all read
from disk to serve request.
What we wanna do is to also add a summary entry after index reaches
a boundary. To deal with oversampling, we want to write 1 byte to
summary for every 2000 bytes written to data file (this will be
eventually made into an option in the config file).
Both conditions must be met to avoid under or oversampling.
That way, the amount of data needed from index file to satify the
request is drastically reduced.
Fixes#1842.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
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 performance.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20170619152629.11703-1-nyh@scylladb.com>
This reverts commit 317d7fc253 (and also the
related 2c57ab84b2). It causes crashes
during range scans, reported by Gleb:
"To reproduce I run SELECT * FROM keyspace1.standard1; on typical c-s
dataset and 3 node cluster.
Backtrace:
at /home/gleb/work/seastar/seastar/core/apply.hh:36
rvalue=<unknown type in /home/gleb/work/seastar/build/release/scylla, CU 0x54cf307, DIE 0x55ebf2a>) at /home/gleb/work/seastar/seastar/core/do_with.hh:57
range=std::vector of length 6, capacity 8 = {...}) at /home/gleb/work/seastar/seastar/core/future-util.hh:142
at ./seastar/core/future.hh:890
at /home/gleb/work/seastar/seastar/core/future-util.hh:119
at /home/gleb/work/seastar/seastar/core/future-util.hh:142
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 performance.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20170614072122.13473-1-nyh@scylladb.com>
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.
That will be needed for optimization that will store decorated keys
in the sstable object, and also for a subsequent work that will
detect wrong metadata (min/max column names) by looking at columns
in the schema. As schema is stored in sstable, there's no longer
a need to store ks and cf names in it.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Remove clustering_key_filter_factory and clustering_key_filtering_context.
Use partition_slice directly with a static get_ranges method.
Signed-off-by: Piotr Jastrzebski <piotr@scylladb.com>
In this unit test, we create using Scylla C++ code, the same large
partition with 13520 CQL rows as we previously imported from Cassandra
for the large partition test. We then verify that the sstable index file
we just wrote is byte-for-byte identical to the one previously created by
Cassandra. They should indeed be identical, because the data file has the
same layout (even if timestamps are different) and our default promoted-
index block size is the same (64K) so the sample of columns should be
identical.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Currently, the main sstable data parsing entry point data_consume_rows()
takes a contiguous range of bytes to read from disk and parse. This range
is supposed to be an entire partition or contiguous group of partitions.
and is self contained (can be parsed without extra information about the
identity of these partitions).
For the promoted index feature (which we will add in a following patch)
we will want the range to span only a part of a partition, and will need
the caller to provide some information not available to the parser (such
as the partition's key). In the future, we will also want to support a
vector of byte ranges, instead of just one.
So in preparation for this, this patch simply replaces the start/end pair
by a new class disk_read_range, which can be easily extended in later
patches. No new functionality is introduced in this patch.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
This patch adds a test that takes an sstable with one partition of 13,520
clustering rows (spanning 700 KB in the data file), and attempts to read
various slices CQL rows, counting that we got back the expected number
of rows.
The sstable included here was generated by Cassandra, and includes a
promoted index. Promoted index reading is not supported yet (we will
add it in the next patch), so for now the code will always read the
entire partition from disk; But still the clustering-key filtering is
already functional, and will drop some of the rows as requested,
so this test will pass.
Later, when we add promoted index support, we should check that this test
still passes - promoted index will make the reads in this test more
efficient (which the test cannot verify), but the important thing to check
is that it doesn't break any of these tests.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
It was noticed that small sstables will accumulate for a column family because
scylla was limited to two compaction per shard, and a column family could have
at most one compaction running at a given shard. With the number of sstables
increasing rapidly, read performance is degraded.
At the moment, our compaction manager works by running two compaction task
handlers that run in parallel to the rest of the system. Each task handler
gets to run when needed, gets a column family from compaction manager queue,
runs compaction on it, and goes to sleep again. That's basically its cycle.
Compaction manager only allows one instance of a column family to be on its
queue, meaning that it's impossible for a column family to be compacted in
parallel. One compaction starts after another for a given column family.
To solve the problem described, we want to concurrently run compaction jobs
of a column family that have different "size tier" (or "weight").
For those unfamiliar, compaction job contains a list of sstables that will be
compacted together.
The "size tier" of a compaction job is the log of the total size of the input
sstables. So a compaction job only gets to run if its "size tier" is not the
same of an ongoing compaction. There is no point in compacting concurrently at
the same "size tier", because that slows down both compactions.
We will no longer queue column families in compaction manager. Instead, we
create a new fiber to run compaction on demand.
This fiber that runs asynchronously will do the following:
1) Get a compaction job from compaction strategy.
2) Calculate "size tier" of compaction job.
3) Run compaction job if its "size tier" is not the same of an ongoing
compaction for the given column family.
As before, it may decide to re-compact a column family based on a stat stored
in column family object.
Ran all compaction-related dtests.
Fixes#1216.
Reviewed-by: Nadav Har'El <nyh@scylladb.com>
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <d30952ff136192a522bde4351926130addec8852.1462311908.git.raphaelsc@scylladb.com>
When we recreate the summary from a missing Summary, we should make
sure it is generated sanely, and that it resembles the Summary that
would have otherwise been there.
In this tests we'll grab one of the Summary tests we've been doing,
and just apply them to the non-existent Summary file. We expect
the same results on those cases. Plus, a new test is added with some
sanity checking.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Vlad and I were working on finding the root of the problems with
refresh. We found that refresh was deleting existing sstable files
because of a bug in a function that was supposed to return the maximum
generation of a column family.
The intention of this function is to get generation from last element
of column_family::_sstables, which is of type std::map.
However, we were incorrectly using std::map::end() to get last element,
so garbage was being read instead of maximum generation.
If the garbage value is lower than the minimum generation of a column
family, then reshuffle_sstables() would set generation of all existing
sstables to a lower value. That would confuse our mechanism used to
delete sstables because sstables loaded at boot stage were touched.
Solution to this problem is about using rbegin() instead of end() to
get last element from column_family::_sstables.
The other problem is that refresh will only load generations that are
larger than or equal to X, so new sstables with lower generation will
not be loaded. Solution is about creating a set with generation of
live SSTables from all shards, and using this set to determine whether
a generation is new or not.
The last change was about providing an unused generation to reshuffle
procedure by adding one to the maximum generation. That's important to
prevent reshuffle from touching an existing SSTable.
Tested 'refresh' under the following scenarios:
1) Existing generations: 1, 2, 3, 4. New ones: 5, 6.
2) Existing generations: 3, 4, 5, 6. New ones: 1, 2.
3) Existing generations: 1, 2, 3, 4. New ones: 7, 8.
4) No existing generation. No new generation.
5) No existing generation. New ones: 1, 2.
I also had to adapt existing testcase for reshuffle procedure.
Fixes#1073.
Signed-off-by: Raphael Carvalho <raphaelsc@scylladb.com>
Message-Id: <1c7b8b7f94163d5cd00d90247598dd7d26442e70.1458694985.git.raphaelsc@scylladb.com>
Currently, if sstable::write_components() is called to write a new sstable
using the same generation of a sstable that exists, a temporary TOC will
be unconditionally created. Afterwards, the same sstable::write_components()
will fail when it reaches sstable::create_data(). The reason is obvious
because data component exists for that generation (in this scenario).
After that, user will not be able to boot scylla anymore because there is
a generation with both a TOC and a temporary TOC. We cannot simply remove a
generation with TOC and temporary TOC because user data will be lost (again,
in this scenario). After all, the temporary TOC was only created because
sstable::write_components() was wrongly called with the generation of a
sstable that exists.
Solution proposed by this patch is to trigger exception if a TOC file
exists for the generation used.
Some SSTable unit tests were also changed to guarantee that we don't try
to overwrite components of an existing sstable.
Refs #1014.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <caffc4e19cdcf25e4c6b9dd277d115422f8246c4.1457643565.git.raphaelsc@scylladb.com>
This patch makes sure that every time we need to create a new generation number -
the very first step in the creation of a new SSTable, the respective CF is already
initialized and populated. Failure to do so can lead to data being overwritten.
Extensive details about why this is important can be found
in Scylla's Github Issue #1014
Nothing should be writing to SSTables before we have the chance to populate the
existing SSTables and calculate what should the next generation number be.
However, if that happens, we want to protect against it in a way that does not
involve overwriting existing tables. This is one of the ways to do it: every
column family starts in an unwriteable state, and when it can finally be written
to, we mark it as writeable.
Note that this *cannot* be a part of add_column_family. That adds a column family
to a db in memory only, and if anybody is about to write to a CF, that was most
likely already called. We need to call this explicitly when we are sure we're ready
to issue disk operations safely.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
All the SSTable read path can now take an io_priority. The public functions will
take a default parameter which is Seastar's default priority.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
We have an API that wraps open_file_dma which we use in some places, but in
many other places we call the reactor version directly.
This patch changes the latter to match the former. It will have the added benefit
of allowing us to make easier changes to these interfaces if needed.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <29296e4ec6f5e84361992028fe3f27adc569f139.1451950408.git.glauber@scylladb.com>
Since bytes is a very generic value that is returned from many calls,
it is easy to pass it by mistake to a function expecting a data_value,
and to get a wrong result. It is impossible for the data_value constructor
to know if the argument is a genuine bytes variable, a data_value of another
type, but serialized, or some other serialized data type.
To prevent misuse, make the data_value(bytes) constructor
(and complementary data_value(optional<bytes>) explicit.
We use boost::any to convert to and from database values (stored in
serlialized form) and native C++ values. boost::any captures information
about the data type (how to copy/move/delete etc.) and stores it inside
the boost::any instance. We later retrieve the real value using
boost::any_cast.
However, data_value (which has a boost::any member) already has type
information as a data_type instance. By teaching data_type intances about
the corresponding native type, we can elimiante the use of boost::any.
While boost::any is evil and eliminating it improves efficiency somewhat,
the real goal is growing native type support in data_type. We will use that
later to store native types in the cache, enabling O(log n) access to
collections, O(1) access to tuples, and more efficient large blob support.
