The sstables replay_position in stats_metadata is
valid only on the originating node and shard.
Therefore, validate the originating host and shard
before using it in compaction or table truncate.
Fixes#10080
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Closesscylladb/scylladb#16550
Fixes some typos as found by codespell run on the code.
In this commit, I was hoping to fix only comments, not user-visible alerts, output, etc.
Follow-up commits will take care of them.
Refs: https://github.com/scylladb/scylladb/issues/16255
Signed-off-by: Yaniv Kaul <yaniv.kaul@scylladb.com>
When off-strategy is disabled, data segregation is not postponed,
meaning that getting partition estimate right is important to
decrease filter's false positives. With streaming, we don't
have min and max timestamps at destination, well, we could have
extended the RPC verb to send them, but turns out we can deduce
easily the amount of windows using default TTL. Given partitioner
random nature, it's not absurd to assume that a given range being
streamed may overlap with all windows, meaning that each range
will yield one sstable for each window when segregating incoming
data. Today, we assume the worst of 100 windows (which is the
max amount of sstables the input data can be segregated into)
due to the lack of metadata for estimating the window count.
But given that users are recommended to target a max of ~20
windows, it means partition estimate is being downsized 5x more
than needed. Let's improve it by using default TTL when
estimating window count, so even on absence of timestamp
metadata, the partition estimation won't be way off.
Fixes#15704.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
default_compaction_progress_monitor returns a reference to a static
object. So, it should be read-only, but its users need to modify it.
Delete default_compaction_progress_monitor and use one's own
compaction_progress_monitor instance where it's needed.
Closesscylladb/scylladb#15800
After "repair: Get rid of the gc_grace_seconds", the sstable's schema (mode,
gc period if applicable, etc) is used to estimate the amount of droppable
data (or determine full expiration = max_deletion_time < gc_before).
It could happen that the user switched from timeout to repair mode, but
sstables will still use the old mode, despite the user asked for a new one.
Another example is when you play with value of grace period, to prevent
data resurrection if repair won't be able to run in a timely manner.
The problem persists until all sstables using old GC settings are recompacted
or node is restarted.
To fix this, we have to feed latest schema into sstable procedures used
for expiration purposes.
Fixes#15643.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closesscylladb/scylladb#15746
compaction_read_monitor_generator is an existing mechanism
for monitoring progress of sstables reading during compaction.
In this change information gathered by compaction_read_monitor_generator
is utilized by task manager compaction tasks of the lowest level,
i.e. compaction executors, to calculate task progress.
compaction_read_monitor_generator has a flag, which decides whether
monitored changes will be registered by compaction_backlog_tracker.
This allows us to pass the generator to all compaction readers without
impacting the backlog.
Task executors have access to compaction_read_monitor_generator_wrapper,
which protects the internals of compaction_read_monitor_generator
and provides only the necessary functionality.
Closesscylladb/scylladb#14878
* github.com:scylladb/scylladb:
compaction: add get_progress method to compaction_task_impl
compaction: find total compaction size
compaction: sstables: monitor validation scrub with compaction_read_generator
compaction: keep compaction_progress_monitor in compaction_task_executor
compaction: use read monitor generator for all compactions
compaction: add compaction_progress_monitor
compaction: add flag to compaction_read_monitor_generator
The estimation assumes that size of other components are irrelevant,
when estimating the number of partitions for each output sstable.
The sstables are split according to the data file size, therefore
size of other files are irrelevant for the estimation.
With certain data models, like single-row partitions containing small
values, the index could be even larger than data.
For example, assume index is as large as data, then the estimation
would say that 2x more sstables will be generated, and as a result,
each sstable are underestimated to have 2x less keys.
Fix it by only accounting size of data file.
Fixes#15726.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closesscylladb/scylladb#15727
Validation scrub bypasses the usual compaction machinery, though it
still needs to be tracked with compaction_progress_monitor so that
we could reach its progress from compaction task executor.
Track sstable scrub in validate mode with read monitors.
Keep compaction_progress_monitor in compaction_task_executor and pass a reference
to it further, so that the compaction progress could be retrieved out of it.
Compaction read monitor generators are used in all compaction types.
Classes which did not use _monitor_generator so far, create it with
_use_backlog_tracker set to no, not to impact backlog tracker.
In the following patches compaction_read_monitor_generator will be used
to find progress of compaction_task_executor's. To avoid unnecessary life
prolongation and exposing internals of the class out of compaction.cc,
compaction_progress_monitor is created.
Compaction class keeps a reference to the compaction_progress_monitor.
Inheriting classes which actually use compaction_read_monitor_generator,
need to set it with set_generator method.
Following patches will use compaction_read_monitor_generator
to track progress of all types of compaction. Some of them should
not be registered in compaction_backlog_tracker.
_use_backlog_tracker flag, which is by default set to true, is
added to compaction_read_monitor_generator and passed to all
compaction_read_monitors created by this generator.
pending_replacement list is used by incremental compaction to
communicate to other ongoing compactions about exhausted sstables
that must be replaced in the sstable set they keep for tombstone
GC purposes.
Reshape doesn't enable tombstone GC, so that list will not
be cleared, which prevents incremental compaction from releasing
sstables referenced by that list. It's not a problem until now
where we want reshape to do incremental compaction.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
That's done by inheriting regular_compaction, which implement
incremental compaction. But reshape still implements its own
methods for creating writer and reader. One reason is that
reshape is not driven by controller, as input sstables to it
live in maintenance set. Another reason is customization
of things like sstable origin, etc.
stop_sstable_writer() is extended because that's used by
regular_compaction to check for possibility of removing
exhausted sstables earlier whenever an output sstable
is sealed.
Also, incremental compaction will be unconditionally
enabled for ICS/LCS during off-strategy.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
That's in preparation to next change that will make reshape
inherit from regular compaction.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
That's needed for enabling incremental compaction to operate, and
needed for subsequent work that enables incremental compaction
for off-strategy, which in turn uses reshape compaction type.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Scylla sstable promises to *never* mutate its input sstables. This
promise was broken by `scylla sstable scrub --scrub-mode=validate`,
because validate moves invalid input sstables into qurantine. This is
unexpected and caused occasional failures in the scrub tests in
test_tools.py. Fix by propagating a flag down to
`scrub_sstables_validate_mode()` in `compaction.cc`, specifying whether
validate should qurantine invalid sstables, then set this flag to false
in `scylla-sstable.cc`. The existing test for validate-mode scrub is
ammended to check that the sstable is not mutated. The test now fails
before the fix and passes afterwards.
Fixes: #14309Closes#15139
Pretty cosmetic change, but it will allow S3 to finally support moving
sstables between states (after this patch it still doesn't)
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
quite a few member variables serves as the configuration for
a given compaction, they are immutable in the life cycle of it,
so for better readability, let's mark them `const`.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#14981
get_compacted_fragments_writer() returns a instance of
`compacted_fragments_writer`, there is no need to cast it again.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#14919
before this change, there are chances that the temporary sstables
created for collecting the GC-able data create by a certain
compaction can be picked up by another compaction job. this
wastes the CPU cycles, adds write amplification, and causes
inefficiency.
in general, these GC-only SSTables are created with the same run id
as those non-GC SSTables, but when a new sstable exhausts input
sstable(s), we proactively replace the old main set with a new one
so that we can free up the space as soon as possible. so the
GC-only SSTables are added to the new main set along with
the non-GC SSTables, but since the former have good chance to
overlap the latter. these GC-only SSTables are assigned with
different run ids. but we fail to register them to the
`compaction_manager` when replacing the main sstable set.
that's why future compactions pick them up when performing compaction,
when the compaction which created them is not yet completed.
so, in this change,
* to prevent sstables in the transient stage from being picked
up by regular compactions, a new interface class is introduced
so that the sstable is always added to registration before
it is added to sstable set, and removed from registration after
it is removed from sstable set. the struct helps to consolidate
the regitration related logic in a single place, and helps to
make it more obvious that the timespan of an sstable in
the registration should cover that in the sstable set.
* use a different run_id for the gc sstable run, as it can
overlap with the output sstable run. the run_id for the
gc sstable run is created only when the gc sstable writer
is created. because the gc sstables is not always created
for all compactions.
please note, all (indirect) callers of
`compaction_task_executor::compact_sstables()` passes a non-empty
`std::function` to this function, so there is no need to check for
empty before calling it. so in this change, the check is dropped.
Fixes#14560
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closes#14725
for faster build times and clear inter-module dependencies, we
should not #includes headers not directly used. instead, we should
only #include the headers directly used by a certain compilation
unit.
in this change, the source files under "/compaction" directories
are checked using clangd, which identifies the cases where we have
an #include which is not directly used. all the #includes identified
by clangd are removed. because some source files rely on the incorrectly
included header file, those ones are updated to #include the header
file they directly use.
if a forward declaration suffice, the declaration is added instead.
see also https://clangd.llvm.org/guides/include-cleaner#unused-include-warning
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Today, SSTable cleanup skips to the next partition, one at a time, when it finds that the current partition is no longer owned by this node.
That's very inefficient because when a cluster is growing in size, existing nodes lose multiple sequential tokens in its owned ranges. Another inefficiency comes from fetching index pages spanning all unowned tokens, which was described in https://github.com/scylladb/scylladb/issues/14317.
To solve both problems, cleanup will now use multi range reader, to guarantee that it will only process the owned data and as a result skip unowned data. This results in cleanup scanning an owned range and then fast forwarding to the next one, until it's done with them all. This reduces significantly the amount of data in the index caching, as index will only be invoked at each range boundary instead.
Without further ado,
before:
`INFO 2023-07-01 07:10:26,281 [shard 0] compaction - [Cleanup keyspace2.standard1 701af580-17f7-11ee-8b85-a479a1a77573] Cleaned 1 sstables to [./tmp/1/keyspace2/standard1-b490ee20179f11ee9134afb16b3e10fd/me-3g7a_0s8o_06uww24drzrroaodpv-big-Data.db:level=0]. 2GB to 1GB (~50% of original) in 26248ms = 81MB/s. ~9443072 total partitions merged to 4750028.`
after:
`INFO 2023-07-01 07:07:52,354 [shard 0] compaction - [Cleanup keyspace2.standard1 199dff90-17f7-11ee-b592-b4f5d81717b9] Cleaned 1 sstables to [./tmp/1/keyspace2/standard1-b490ee20179f11ee9134afb16b3e10fd/me-3g7a_0s4m_5hehd2rejj8w15d2nt-big-Data.db:level=0]. 2GB to 1GB (~50% of original) in 17424ms = 123MB/s. ~9443072 total partitions merged to 4750028.`
Fixes#12998.
Fixes#14317.
Closes#14469
* github.com:scylladb/scylladb:
test: Extend cleanup correctness test to cover more cases
compaction: Make SSTable cleanup more efficient by fast forwarding to next owned range
sstables: Close SSTable reader if index exhaustion is detected in fast forward call
sstables: Simplify sstable reader initialization
compaction: Extend make_sstable_reader() interface to work with mutation_source
test: Extend sstable partition skipping test to cover fast forward using token
Today, SSTable cleanup skips to the next partition, one at a time, when it finds
that the current partition is no longer owned by this node.
That's very inefficient because when a cluster is growing in size, existing
nodes lose multiple sequential tokens in its owned ranges. Another inefficiency
comes from fetching index pages spanning all unowned tokens, which was described
in #14317.
To solve both problems, cleanup will now use multi range reader, to guarantee
that it will only process the owned data and as a result skip unowned data.
This results in cleanup scanning an owned range and then fast forwarding to the
next one, until it's done with them all. This reduces significantly the amount
of data in the index caching, as index will only be invoked at each range
boundary instead.
Without further ado,
before:
... 2GB to 1GB (~50% of original) in 26248ms = 81MB/s. ~9443072 total partitions merged to 4750028.
after:
... 2GB to 1GB (~50% of original) in 17424ms = 123MB/s. ~9443072 total partitions merged to 4750028.
Fixes#12998.
Fixes#14317.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Today, we base compaction throughput on the amount of data written,
but it should be based on the amount of input data compacted
instead, to show the amount of data compaction had to process
during its execution.
A good example is a compaction which expire 99% of data, and
today throughput would be calculated on the 1% written, which
will mislead the reader to think that compaction was terribly
slow.
Fixes#14533.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closes#14615
As the goal is to make compaction filter to the next owned range,
make_sstable_reader() should be extended to create a reader with
parameters forwarded from mutation_source interface, which will
be used when wiring cleanup with multi range reader.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
In that level no io_priority_class-es exist. Instead, all the IO happens
in the context of current sched-group. File API no longer accepts prio
class argument (and makes io_intent arg mandatory to impls).
So the change consists of
- removing all usage of io_priority_class
- patching file_impl's inheritants to updated API
- priority manager goes away altogether
- IO bandwidth update is performed on respective sched group
- tune-up scylla-gdb.py io_queues command
The first change is huge and was made semi-autimatically by:
- grep io_priority_class | default_priority_class
- remove all calls, found methods' args and class' fields
Patching file_impl-s is smaller, but also mechanical:
- replace io_priority_class& argument with io_intent* one
- pass intent to lower file (if applicatble)
Dropping the priority manager is:
- git-rm .cc and .hh
- sed out all the #include-s
- fix configure.py and cmakefile
The scylla-gdb.py update is a bit hairry -- it needs to use task queues
list for IO classes names and shares, but to detect it should it checks
for the "commitlog" group is present.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Closes#13963
Commit 8c4b5e4283 introduced an optimization which only
calculates max purgeable timestamp when a tombstone satisfy the
grace period.
Commit 'repair: Get rid of the gc_grace_seconds' inverted the order,
probably under the assumption that getting grace period can be
more expensive than calculating max purgeable, as repair-mode GC
will look up into history data in order to calculate gc_before.
This caused a significant regression on tombstone heavy compactions,
where most of tombstones are still newer than grace period.
A compaction which used to take 5s, now takes 35s. 7x slower.
The reason is simple, now calculation of max purgeable happens
for every single tombstone (once for each key), even the ones that
cannot be GC'ed yet. And each calculation has to iterate through
(i.e. check the bloom filter of) every single sstable that doesn't
participate in compaction.
Flame graph makes it very clear that bloom filter is a heavy path
without the optimization:
45.64% 45.64% sstable_compact sstable_compaction_test_g
[.] utils::filter::bloom_filter::is_present
With its resurrection, the problem is gone.
This scenario can easily happen, e.g. after a deletion burst, and
tombstones becoming only GC'able after they reach upper tiers in
the LSM tree.
Before this patch, a compaction can be estimated to have this # of
filter checks:
(# of keys containing *any* tombstone) * (# of uncompacting sstable
runs[1])
[1] It's # of *runs*, as each key tend to overlap with only one
fragment of each run.
After this patch, the estimation becomes:
(# of keys containing a GC'able tombstone) * (# of uncompacting
runs).
With repair mode for tombstone GC, the assumption, that retrieval
of gc_before is more expensive than calculating max purgeable,
is kept. We can revisit it later. But the default mode, which
is the "timeout" (i.e. gc_grace_seconds) one, we still benefit
from the optimization of deferring the calculation until
needed.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closes#13908
with off-strategy, input list size can be close to 1k, which will
lead to unneeded reallocations when formatting the list for
logging.
in the past, we faced stalls in this area, and excessive reallocation
(log2 ~1k = ~10) may have contributed to that.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closes#13907
If tombstone GC was disabled, compaction will ensure that fully expired
sstables won't be bypassed and that no expired tombstones will be
purged. Changing the value takes immediate effect even on ongoing
compactions.
Not wired into an API yet.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Currently said method creates a combined reader from all the sstables
passed to it then validates this combined reader.
Change it to validate each sstable (reader) individually in preparation
of the new validate method which can handle a single sstable at a time.
Note that this is not going to make much impact in practice, all callers
pass a single sstable to this method already.
Now, with f1bbf705f9
(Cleanup sstables in resharding and other compaction types),
we may filter sstables as part of resharding compaction
and the assertion that all tokens are owned by the current
shard when filtering is no longer true.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Refactor the printing logic in compaction::formatted_sstables_list
out to sstables::to_string(const shared_sstable&, bool include_origin)
and operator<<(const shared_sstable) on top of it.
So that we can easily print std::vector<shared_sstable>
from compaction_manager in the next patch.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Move the token filtering logic down from cleanup_compaction
to regular_compaction and class compaction so it can be
reused by other compaction types.
Create a _owned_ranges_checker in class compaction
when _owned_ranges is engaged, and use it in
compaction::setup to filter partitions based on the owned ranges.
Ref scylladb/scylladb#12998
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Move the owned_ranges_ptr, currently used only by
cleanup and upgrade compactions, to the generic
compaction descriptor so we apply cleanup in other
compaction types.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
The call moves the sstable to the specified state.
The change state is translated into the storage driver state change
which is for todays filesystem storage means moving between directories.
The "normal" state maps to the base dir of the table, there's no
dedicated subdir for this state and this brings some trouble into the
play.
The thing is that in order to check if an sstable is in "normal" state
already its impossible to compare filename of its path to any
pre-defined values, as tables' basdirs are dynamic. To overcome this,
the change-state call checks that the sstable is in one of "known"
sub-states, and assumes that it's in normal state otherwise.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
these warnings are found by Clang-17 after removing
`-Wno-unused-lambda-capture` and '-Wno-unused-variable' from
the list of disabled warnings in `configure.py`.
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Schema related files are moved there. This excludes schema files that
also interact with mutations, because the mutation module depends on
the schema. Those files will have to go into a separate module.
Closes#12858
Move mutation-related files to a new mutation/ directory. The names
are kept in the global namespace to reduce churn; the names are
unambiguous in any case.
mutation_reader remains in the readers/ module.
mutation_partition_v2.cc was missing from CMakeLists.txt; it's added in this
patch.
This is a step forward towards librarization or modularization of the
source base.
Closes#12788
Its _it member keeps state about the current range.
Although it's modified by the method, this is an implementation
detail that irrelevant to the caller, hence mark the
belongs_to_current_node method as const (and noexcept while
at it).
This allows the caller, cleanup_compaction, to use it from
inside a const method, without having to mark
its respective member as mutable too.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Closes#12634
This fixes a long standing bug related to handling of non-full
clustering keys, issue #1446.
after_key() was creating a position which is after all keys prefixed
by a non-full key, rather than a position which is right after that
key.
This will issue will be caught by cql_query_test::test_compact_storage
in debug mode when mutation_partition_v2 merging starts inserting
sentinels at position after_key() on preemption.
It probably already causes problems for such keys.
