Today, compaction can act much more aggressive than it really has to, because
the strategy and its definition of backlog are completely decoupled.
The backlog definition for size-tiered, which is inherited by all
strategies (e.g.: LCS L0, TWCS' windows), is built on the assumption that the
world must reach the state of zero amplification. But that's unrealistic and
goes against the intent amplification defined by the compaction strategy.
For example, size tiered is a write oriented strategy which allows for extra
space amplification for compaction to keep up with the high write rate.
It can be seen today, in many deployments, that compaction shares is either
close to 1000, or even stuck at 1000, even though there's nothing to be done,
i.e. the compaction strategy is completely satisfied.
When there's a single sstable per tier, for example.
This means that whenever a new compaction job kicks in, it will act much more
aggressive because of the high shares, caused by false backlog of the existing
tables. This translates into higher P99 latencies and reduced throughput.
Solution
========
This problem can be fixed, as proposed in the document "Fixing compaction
aggressiveness due to suboptimal definition of zero backlog by controller" [1],
by removing backlog of tiers that don't have to be compacted now, like a tier
that has a single file. That's about coupling the strategy goal with the
backlog definition. So once strategy becomes satisfied, so will the controller.
Low-efficiency compaction, like compacting 2 files only or cross-tier, only
happens when system is under little load and can proceed at a slower pace.
Once efficient jobs show up, ongoing compactions, even if inefficient, will get
more shares (as efficient jobs add to the backlog) so compaction won't fall
behind.
With this approach, throughput and latency is improved as cpu time is no longer
stolen (unnecessarily) from the foreground requests.
[1]: https://docs.google.com/document/d/1EQnXXGWg6z7VAwI4u8AaUX1vFduClaf6WOMt2wem5oQFixes#4588.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
This new interface allows table to communicate multiple changes in the
SSTable set with a single call, which is useful on compaction completion
for example.
With this new interface, the size tiered backlog tracker will be able to
know when compaction completed, which will allow it to recompute tiers
and their backlog contribution, if any. Without it, tiered tracker
would have to recompute tiers for every change, which would be terribly
expensive.
The old remove/add interface are being removed in favor of the new one.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
If the tracker is disabled, we never get to access the underlying
implementation anymore. It makes sense to clear _impl on
disable(). So table::stop() can call its backlog tracker's disable
method, clearing all its state. This is important for clean
shutdown, as any sstable in tracker state may cause sstable
manager to hang when being stopped.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Normalized backlog metric is important for understanding the controller
behavior as the controller acts on normalized backlog for yielding an
output, not the raw backlog value in bytes.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Use exponential_backoff_retry::retry(abort_source&)
when sleeping between retries and request abort
when the task is stopped.
Fixes#10112
Test: unit(dev)
Closes#10113
* github.com:scylladb/scylla:
compaction_manager: allow stopping sleeping tasks
compaction_manager: task: add make_compaction_stopped_exception
compaction_manager: task: refactor stop
Use exponential_backoff_retry::retry(abort_source&)
when sleeping between retries and request abort
when the task is stopped.
Fixes#10112
Test: unit(dev)
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Provide a function to make a sstables::compaction_stopped_exception
based on the information in the stopped task.
To be reused by the next patch that will
also throw this exception from the retry sleep path,
when the task is stopped.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Before we add a v2 output option to the compactor, we want to get rid of
all the v1 inputs to make it simpler. This means that for a while the
compacting reader will be in a strange place of having a v2 input and a
v1 output. Hopefully, not for long.
This reverts commit 4c05e5f966.
Moving cleanup to maintenance group made its operation time up to
10x slower than previous release. It's a blocker to 4.6 release,
so let's revert it until we figure this all out.
Probably this happens because maintenance group is fixed at a
relatively small constant, and cleanup may be incrementally
generating backlog for regular compaction, where the former is
fighting for resources against the latter.
Fixes#10060.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20220213184306.91585-1-raphaelsc@scylladb.com>
Now that the table layer is using perform_offstrategy,
submit_offstrategy is no longer in use and can be deleted.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Return a future from perform_offstrategy, resolved
when the offstrategy compaction completes so that callers
can wait on it.
submit_offstrategy still submits the offstrategy compaction
in the background.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Instead of lengthy blurbs, switch to single-line, machine-readable
standardized (https://spdx.dev) license identifiers. The Linux kernel
switched long ago, so there is strong precedent.
Three cases are handled: AGPL-only, Apache-only, and dual licensed.
For the latter case, I chose (AGPL-3.0-or-later and Apache-2.0),
reasoning that our changes are extensive enough to apply our license.
The changes we applied mechanically with a script, except to
licenses/README.md.
Closes#9937
Currently, when TWCS reshape finds a bucket containing more than 32
files, it will blindly resize that bucket to 32.
That's very bad because it doesn't take into consideration that
compaction efficiency depends on relative sizes of files being
compacted together, meaning that a huge file can be compacted with
a tiny one, producing lots of write amplification.
To solve this problem, STCS reshape logic will now be reused in
each time bucket. So only similar-sized files are compacted together
and the time bucket will be considered reshaped once its size tiers
are properly compacted, according to the reshape mode.
Fixes#9938.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20220117205000.121614-1-raphaelsc@scylladb.com>
"
With this series the mutation compactor can now consume a v2 stream. On
the output side it still uses v1, so it can now act as an online
v2->v1 converter. This allows us to push out v2->v1 conversion to as far
as the compactor, usually the next to last component in a read pipeline,
just before the final consumer. For reads this is as far as we can go,
as the intra-node ABI and hence the result-sets built are v1. For
compaction we could go further and eliminate conversion altogether, but
this requires some further work on both the compactor and the sstable
writer and so it is left to be done later.
To summarize, this patchset enables a v2 input for the compactor and it
updates compaction and single partition reads to use it.
"
* 'mutation-compactor-consume-v2/v1' of https://github.com/denesb/scylla:
table: add make_reader_v2()
querier: convert querier_cache and {data,mutation}_querier to v2
compaction: upgrade compaction::make_interposer_consumer() to v2
mutation_reader: remove unecessary stable_flattened_mutations_consumer
compaction/compaction_strategy: convert make_interposer_consumer() to v2
mutation_writer: migrate timestamp_based_splitting_writer to v2
mutation_writer: migrate shard_based_splitting_writer to v2
mutation_writer: add v2 clone of feed_writer and bucket_writer
flat_mutation_reader_v2: add reader_consumer_v2 typedef
mutation_reader: add v2 clone of queue_reader
compact_mutation: make start_new_page() independent of mutation_fragment version
compact_mutation: add support for consuming a v2 stream
compact_mutation: extract range tombstone consumption into own method
range_tombstone_assembler: add get_range_tombstone_change()
range_tombstone_assembler: add get_current_tombstone()
run_with_compaction_disabled() is used to temporarily disable compaction
for a table T. Not only regular compaction, but all types.
Turns out it's stopping all types but it's only preventing new regular
compactions from starting. So major for example can start even with
compaction temporarily disabled. This is fixed by not allowing
compaction of any type if disabled. This wasn't possible before as
scrub incorrectly ran entirely with compaction disabled, so it wouldn't
be able to start, but now it only disables compaction while retrieving
its candidate list.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20220107154942.59800-1-raphaelsc@scylladb.com>
Said wrapper was conceived to make unmovable `compact_mutation` because
readers wanted movable consumers. But `compact_mutation` is movable for
years now, as all its unmovable bits were moved into an
`lw_shared_ptr<>` member. So drop this unnecessary wrapper and its
unnecessary usages.
Move the ::database, ::keyspace, and ::table classes to a new replica
namespace and replica/ directory. This designates objects that only
have meaning on a replica and should not be used on a coordinator
(but note that not all replica-only classes should be in this module,
for example compaction and sstables are lower-level objects that
deserve their own modules).
The module is imperfect - some additional classes like distributed_loader
should also be moved, but there is only one way to untie Gordian knots.
Closes#9872
* github.com:scylladb/scylla:
replica: move ::database, ::keyspace, and ::table to replica namespace
database: Move database, keyspace, table classes to replica/ directory
Move replica-oriented classes to the replica namespace. The main
classes moved are ::database, ::keyspace, and ::table, but a few
ancillary classes are also moved. There are certainly classes that
should be moved but aren't (like distributed_loader) but we have
to start somewhere.
References are adjusted treewide. In many cases, it is obvious that
a call site should not access the replica (but the data_dictionary
instead), but that is left for separate work.
scylla-gdb.py is adjusted to look for both the new and old names.
Tables waiting for a chance to run reshape wouldn't trigger stop
exception, as the exception was only being triggered for ongoing
compactions. Given that stop reshape API must abort all ongoing
tasks and all pending ones, let's change run_custom_job() to
trigger the exception if it found that the pending task was
asked to stop.
Tests:
dtest: compaction_additional_test.py::TestCompactionAdditional::test_stop_reshape_with_multiple_keyspaces
unit: dev
Fixes#9836.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <20211223002157.215571-1-raphaelsc@scylladb.com>
The database, keyspace, and table classes represent the replica-only
part of the objects after which they are named. Reading from a table
doesn't give you the full data, just the replica's view, and it is not
consistent since reconciliation is applied on the coordinator.
As a first step in acknowledging this, move the related files to
a replica/ subdirectory.
"
Completes coroutinization of rewrite_sstables().
tests: UNIT(debug)
"
* 'rewrite_sstable_coroutinization' of https://github.com/raphaelsc/scylla:
compaction_manager: coroutinize main loop in sstable rewrite procedure
compaction_manager: coroutinize exception handling in sstable rewrite procedure
compaction_manager: mark task::finish_compaction() as noexcept
compaction_manager: make maybe_stop_on_error() more flexible
We can stop only specific compaction types.
Reshard should be excluded since it mustn't be stopped.
And other types of compaction types like "VALIDATION" or "INDEX_BUILD"
are valid in terms of their syntax but unsupported by scylla so we better
return an error rather than appear to support them.
Test: unit(dev)
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20211222133449.2177746-1-bhalevy@scylladb.com>
It's hard to integrate maybe_stop_on_error() with coroutines as it
accepts a resolved future, not an exception pointer. Let's adjust
its interface, making it more flexible to work with.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The gc_grace_seconds is a very fragile and broken design inherited from
Cassandra. Deleted data can be resurrected if cluster wide repair is not
performed within gc_grace_seconds. This design pushes the job of making
the database consistency to the user. In practice, it is very hard to
guarantee repair is performed within gc_grace_seconds all the time. For
example, repair workload has the lowest priority in the system which can
be slowed down by the higher priority workload, so that there is no
guarantee when a repair can finish. A gc_grace_seconds value that is
used to work might not work after data volume grows in a cluster. Users
might want to avoid running repair during a specific period where
latency is the top priority for their business.
To solve this problem, an automatic mechanism to protect data
resurrection is proposed and implemented. The main idea is to remove the
tombstone only after the range that covers the tombstone is repaired.
In this patch, a new table option tombstone_gc is added. The option is
used to configure tombstone gc mode. For example:
1) GC a tombstone after gc_grace_seconds
cqlsh> ALTER TABLE ks.cf WITH tombstone_gc = {'mode':'timeout'} ;
This is the default mode. If no tombstone_gc option is specified by the
user. The old gc_grace_seconds based gc will be used.
2) Never GC a tombstone
cqlsh> ALTER TABLE ks.cf WITH tombstone_gc = {'mode':'disabled'};
3) GC a tombstone immediately
cqlsh> ALTER TABLE ks.cf WITH tombstone_gc = {'mode':'immediate'};
4) GC a tombstone after repair
cqlsh> ALTER TABLE ks.cf WITH tombstone_gc = {'mode':'repair'};
In addition to the 'mode' option, another option 'propagation_delay_in_seconds'
is added. It defines the max time a write could possibly delay before it
eventually arrives at a node.
A new gossip feature TOMBSTONE_GC_OPTIONS is added. The new tombstone_gc
option can only be used after the whole cluster supports the new
feature. A mixed cluster works with no problem.
Tests: compaction_test.py, ninja test
Fixes#3560
[avi: resolve conflicts vs data_dictionary]
"
TWCS perform STCS on a window as long as it's the most recent one.
From there on, TWCS will compact all files in the past window into
a single file. With some moderate write load, it could happen that
there's still some compaction activity in that past window, meaning
that per-window major may miss some files being currently compacted.
As a result, a past window may contain more than 1 file after all
compaction activity is done on its behalf, which may increase read
amplification. To avoid that, TWCS will now make sure that per-window
major is serialized, to make sure no files are missed.
Fixes#9553.
tests: unit(dev).
"
* 'fix_twcs_per_window_major_v3' of https://github.com/raphaelsc/scylla:
TWCS: Make sure major on past window is done on all its sstables
TWCS: remove needless param for STCS options
TWCS: kill unused param in newest_bucket()
compaction: Implement strategy control and wire it
compaction: Add interface to control strategy behavior.
Allow stopping compaction by type on a given keyspace and list of tables.
Also add api unit test suite that tests the existing `stop_compaction` api
and the new `stop_keyspace_compaction` api.
Fixes#9700Closes#9746
* github.com:scylladb/scylla:
api: storage_service: validate_keyspace: improve exception error message
api: compaction_manager: add stop_keyspace_compaction
api: storage_service: expose validate_keyspace and parse_tables
api: compaction_manager: stop_compaction: fix type description
compaction_manager: stop_compaction: expose optional table*
test: api: add basic compaction_manager test
Currently when scrub/validate is stopped (e.g. via the api),
scrub_validate_mode_validate_reader co_return:s without
closing the reader passed to it - causing a crash due
to internal error check, see #9766.
Throwing a compaction_stopped_exception rather than co_return:ing
an exception will be handled as any other exeption, including closing
the reader.
Fixes#9766
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20211213125528.2422745-1-bhalevy@scylladb.com>
Once current window is sealed, TWCS is supposed to compact all its
sstables into one. If there's ongoing compaction, it can happen that
sstables are missed and therefore past windows will contain more than
one sstable. Additionally, it could happen that major doesn't happen
at all if under heavy load. All these problems are fixed by serializing
major on past window and also postponing it if manager refuses to run
the job now.
Fixes#9553.
Reviewed-by: Benny Halevy <bhalevy@scylladb.com>
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
STCS option can be retrieved from class member, as newest_bucket()
is no longer a static function. let's get rid of it.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
