Introduces support to split large partitions during compaction. Today, compaction can only split input data at partition boundary, so a large partition is stored in a single file. But that can cause many problems, like memory pressure (e.g.: https://github.com/scylladb/scylladb/issues/4217), and incremental compaction can also not fulfill its promise as the file storing the large partition can only be released once exhausted.
The first step was to add clustering range metadata for first and last partition keys (retrieved from promoted index), which is crucial to determine disjointness at clustering level, and also the order at which the disjoint files should be opened for incremental reading.
The second step was to extend sstable_run to look at clustering dimension, so a set of files storing disjoint ranges for the same partition can live in the same sstable run.
The final step was to introduce the option for compaction to split large partition being written if it has exceeded the size threshold.
What's next? Following this series, a reader will be implemented for sstable_run that will incrementally open the readers. It can be safely built on the assumption of the disjoint invariant after the second step aforementioned.
Closes#11233
* github.com:scylladb/scylladb:
test: Add test for large partition splitting on compaction
compaction: Add support to split large partitions
sstable: Extend sstable_run to allow disjointness on the clustering level
sstables: simplify will_introduce_overlapping()
test: move sstable_run_disjoint_invariant_test into sstable_datafile_test
test: lib: Fix inefficient merging of mutations in make_sstable_containing()
sstables: Keep track of first partition's first pos and last partition's last pos
sstables: Rename min/max position_range to a descriptive name
sstables_manager: Add sstable metadata reader concurrency semaphore
sstables: Add ability to find first or last position in a partition
Let's introduce a reader_concurrency_semaphore for reading sstable
metadata, to avoid an OOM due to unlimited concurrency.
The concurrency on startup is not controlled, so it's important
to enforce a limit on the amount of memory used by the parallel
readers.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The intention was for these logs to be printed during the
database shutdown sequence, but it was overlooked that it's not
the only place where commitlog::shutdown is called.
Commitlogs are started and shut down periodically by hinted handoff.
When that happens, these messages spam the log.
Fix that by adding INFO commitlog shutdown logs to database::stop,
and change the level of the commitlog::shutdown log call to DEBUG.
Fixes#11508Closes#11536
Now memtables live in compaction_group. Also introduced function
that selects group based on token, but today table always return
the single group managed by it. Once multiple groups are supported,
then the function should interpret token content to select the
group.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The group is now responsible for providing the compound set.
table still has one compound set, which will span all groups for
the cases we want to ignore the group isolation.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
This commit is restricted to moving maintenance set into compaction_group.
Next, we'll introduce compound set into it.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
This commit is restricted to moving main set into compaction_group.
Next, we'll move maintenance set into it and finally the memtable.
A method is introduced to figure out which group a sstable belongs
to, but it's still unimplemented as table is still limited to
a single group.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Compaction group is a new abstraction used to group SSTables
that are eligible to be compacted together. By this definition,
a table in a given shard has a single compaction group.
The problem with this approach is that data from different vnodes
is intermixed in the same sstable, making it hard to move data
in a given sstable around.
Therefore, we'll want to have multiple groups per table.
A group can be thought of an isolated LSM tree where its memtable
and sstable files are isolated from other groups.
As for the implementation, the idea is to take a very incremental
approach.
In this commit, we're introducing a single compaction group to
table.
Next, we'll migrate sstable and maintenance set from table
into that single compaction group. And finally, the memtable.
Cache will be shared among the groups, for simplicity.
It works due to its ability to invalidate a subset of the
token range.
There will be 1:1 relationship between compaction_group and
table_state.
We can later rename table_state to compaction_group_state.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
await_pending_ops() is today marked noexcept, so doesn't have to
be implemented with finally() semantics.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
and override it in table::table_state to get the tombstone_gc_state
from the table's compaction_manager.
It is going to be used in the next patched to pass the gc state
from the compaction_strategy down to sstables and compaction.
table_state_for_test was modified to just keep a null
tombstone_gc_state.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
To be used by generate_update() for getting the
tombstone_gc_state via the table's compaction_manager.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
and use it to access the repair history maps.
At this introductory patch, we use default-constructed
tombstone_gc_state to access the thread-local maps
temporarily and those use sites will be replaced
in following patches that will gradually pass
the tombstone_gc_state down from the compaction_manager
to where it's used.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
drop_repair_history_map_for_table is called on each shard
when database::truncate is done, and the table is stopped.
dropping it before the table is stopped is too early.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
on_compaction_completion() is not very descriptive. let's rename
it, following the example of
update_sstable_lists_on_off_strategy_completion().
Also let's coroutinize it, so to remove the restriction of running
it inside a thread only.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closes#11407
Currently the state of LSA is scattered across a handful of global variables. This series consolidates all these into a single one: the shard tracker. Beyond reducing the number of globals (the less globals, the better) this paves the way for a planned de-globalization of the shard tracker itself.
There is one separate global left, the static migrators registry. This is left as-is for now.
Closes#11284
* github.com:scylladb/scylladb:
utils/logalloc: remove reclaim_timer:: globals
utils/logalloc: make s_sanitizer_report_backtrace global a member of tracker
utils/logalloc: tracker_reclaimer_lock: get shard tracker via constructor arg
utils/logalloc: move global stat accessors to tracker
utils/logalloc: allocating_section: don't use the global tracker
utils/logalloc: pass down tracker::impl reference to segment_pool
utils/logalloc: move segment pool into tracker
utils/logalloc: add tracker member to basic_region_impl
utils/logalloc: make segment independent of segment pool
Aborting too soon on ENOSPC is too harsh, leading to loss of
availability of the node for reads, while restarting it won't
solve the ENOSPC condition.
Fixes#11245
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Closes#11246
Fixes#11349
In 7396de7 (and refactorings before it) the set of prioritized keyspaces (and processing thereof)
was removed, due to apparent non-usage (which is true for open-source version).
This functionality is however required for certain features of the enterprise version (ear).
As such is needs to be restored and reenabled. This patch and revert before it does so, adapted
to the recent version of this file.
This reverts commit 7396de72b1.
In 7396de7 (and refactorings before it) the set of prioritized keyspaces (and processing thereof)
was removed, due to apparent non-usage (which is true for open-source version).
This functionality is however required for certain features of the enterprise version (ear).
As such is needs to be restored and reenabled. This reverts the actual commit, patch after
ensures we use the prio set.
These are pretend free functions, accessing globals in the background,
make them a member of the tracker instead, which everything needed
locally to compute them. Callers still have to access these stats
through the global tracker instance, but this can be changed to happen
through a local instance. Soon....
Instead of querier_cache::evict_all_for_table(). The new method cover
all queriers and in addition any other inactive reads registered on the
semaphore. In theory by the time we detach a table, no regular inactive
reads should be in the semaphore anymore, but if there is any still, we
better evict them before the table is destroyed, they might attempt to
access it in when destroyed later.
Many tombstones in a partition is a problem that has been plaguing queries since the inception of Scylla (and even before that as they are a pain in Apache Cassandra too). Tombstones don't count towards the query's page limit, neither the size nor the row number one. Hence, large spans of tombstones (be that row- or range-tombstones) are problematic: the query can time out while processing this span of tombstones, as it waits for more live rows to fill the page. In the extreme case a partition becomes entirely unreadable, all read attempts timing out, until compaction manages to purge the tombstones.
The solution proposed in this PR is to pass down a tombstone limit to replicas: when this limit is reached, the replica cuts the page and marks it as short one, even if the page is empty currently. To make this work, we use the last-position infrastructure added recently by 3131cbea62, so that replicas can provide the position of the last processed item to continue the next page from. Without this no forward progress could be made in the case of an empty page: the query would continue from the same position on the next page, having to process the same span of tombstones.
The limit can be configured with the newly added `query_tombstone_limit` configuration item, defaulted to 10000. The coordinator will pass this to the newly added `tombstone_limit` field of `read_command`, if the `replica_empty_pages` cluster feature is set.
Upgrade sanity test was conducted as following:
* Created cluster of 3 nodes with RF=3 with master version
* Wrote small dataset of 1000 rows.
* Deleted prefix of 980 rows.
* Started read workload: `scylla-bench -mode=read -workload=uniform -replication-factor=3 -nodes 127.0.0.1,127.0.0.2,127.0.0.3 -clustering-row-count=10000 -duration=10m -rows-per-request=9000 -page-size=100`
* Also did some manual queries via `cqlsh` with smaller page size and tracing on.
* Stopped and upgraded each node one-by-one. New nodes were started by `--query-tombstone-page-limit=10`.
* Confirmed there are no errors or read-repairs.
Perf regression test:
```
build/release/test/perf/perf_simple_query_g -c1 -m2G --concurrency=1000 --task-quota-ms 10 --duration=60
```
Before:
```
median 133665.96 tps ( 62.0 allocs/op, 12.0 tasks/op, 43007 insns/op, 0 errors)
median absolute deviation: 973.40
maximum: 135511.63
minimum: 104978.74
```
After:
```
median 129984.90 tps ( 62.0 allocs/op, 12.0 tasks/op, 43181 insns/op, 0 errors)
median absolute deviation: 2979.13
maximum: 134538.13
minimum: 114688.07
```
Diff: +~200 instruction/op.
Fixes: https://github.com/scylladb/scylla/issues/7689
Fixes: https://github.com/scylladb/scylla/issues/3914
Fixes: https://github.com/scylladb/scylla/issues/7933
Refs: https://github.com/scylladb/scylla/issues/3672Closes#11053
* github.com:scylladb/scylladb:
test/cql-pytest: add test for query tombstone page limit
query-result-writer: stop when tombstone-limit is reached
service/pager: prepare for empty pages
service/storage_proxy: set smallest continue pos as query's continue pos
service/storage_proxy: propagate last position on digest reads
query: result_merger::get() don't reset last-pos on short-reads and last pages
query: add tombstone-limit to read-command
service/storage_proxy: add get_tombstone_limit()
query: add tombstone_limit type
db/config: add config item for query tombstone limit
gms: add cluster feature for empty replica pages
tree: don't use query::read_command's IDL constructor
The query result writer now counts tombstones and cuts the page (marking
it as a short one) when the tombstone limit is reached. This is to avoid
timing out on large span of tombstones, especially prefixes.
In the case of unpaged queries, we fail the read instead, similarly to
how we do with max result size.
If the limit is 0, the previous behaviour is used: tombstones are not
taken into consideration at all.
This series converts the synchronous `effective_replication_map::get_range_addresses` to async
by calling the replication strategy async entry point with the same name, as its callers are already async
or can be made so easily.
To allow it to yield and work on a coherent view of the token_metadata / topology / replication_map,
let the callers of this patch hold a effective_replication_map per keyspace and pass it down
to the (now asynchronous) functions that use it (making affected storage_service methods static where possible
if they no longer depend on the storage_service instance).
Also, the repeated calls to everywhere_replication_strategy::calculate_natural_endpoints
are optimized in this series by introducing a virtual abstract_replication_strategy::has_static_natural_endpoints predicate
that is true for local_strategy and everywhere_replication_strategy, and is false otherwise.
With it, functions repeatedly calling calculate_natural_endpoints in a loop, for every token, will call it only once since it will return the same result every time anyhow.
Refs #11005
Doesn't fix the issue as the large allocation still remains until we make change dht::token_range_vector chunked (chunked_vector cannot be used as is at the moment since we require the ability to push also to the front when unwrapping)
Closes#11009
* github.com:scylladb/scylladb:
effective_replication_map: make get_range_addresses asynchronous
range_streamer: add_ranges and friends: get erm as param
storage_service: get_new_source_ranges: get erm as param
storage_service: get_changed_ranges_for_leaving: get erm as param
storage_service: get_ranges_for_endpoint: get erm as param
repair: use get_non_local_strategy_keyspaces_erms
database: add get_non_local_strategy_keyspaces_erms
database: add get_non_local_strategy_keyspaces
storage_service: coroutinize update_pending_ranges
effective_replication_map: add get_replication_strategy
effective_replication_map: get_range_addresses: use the precalculated replication_map
abstract_replication_strategy: get_pending_address_ranges: prevent extra vector copies
abstract_replication_strategy: reindent
utils: sequenced_set: expose set and `contains` method
abstract_replication_strategy: calculate_natural_endpoints: return endpoint_set
utils: sequenced_set: templatize VectorType
utils: sanitize sequenced_set
utils: sequenced_set: delete mutable get_vector method
This reverts commit bcadd8229b, reversing
changes made to cf528d7df9. Since
4bd4aa2e88 ("Merge 'memtable, cache: Eagerly
compact data with tombstones' from Tomasz Grabiec"), memtable is
self-compacting and the extra compaction step only reduces throughput.
The unit test in memtable_test.cc is not reverted as proof that the
revert does not cause a regression.
Closes#11243
To be used by coordinator side code to determine the correct tombstone
limit to pass to read-command (tombstone limit field added in the next
commit). When this limit is non-zero, the replica will start cutting
pages after the tombstone limit is surpassed.
This getter works similarly to `get_max_result_size()`: if the cluster
feature for empty replica pages is set, it will return the value
configured via db::config::query_tombstone_limit. System queries always
use a limit of 0 (unlimited tombstones).
To be used for getting a coheret set of all keyspaces
with non-local replication strategy and their respective
effective_replication_map.
As an example, use it in this patch in
storage_service::update_pending_ranges.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
For node operations, we currently call get_non_system_keyspaces
but really want to work on all keyspace that have non-local
replication strategy as they are replicated on other nodes.
Reflect that in the replica::database function name.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Define table_schema_version as a distinct tagged_uuid class,
So it can be differentiated from other uuid-class types,
in particular table_id.
Added reversed(table_schema_version) for convenience
and uniformity since the same logic is currently open coded
in several places.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Define table_id as a distinct utils::tagged_uuid modeled after raft
tagged_id, so it can be differentiated from other uuid-class types,
in particular from table_schema_version.
Fixes#11207
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Add maybe_yield calls in tight loop, potentially
over thousands of sstable names to prevent reactor stalls.
Although the per-sstable cost is very small, we've experienced
stalls realted to printing in O(#sstables) in compaction.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Handle exceptions, making sure the output
stream is properly closed in all cases,
and an intermediate error, if any, is returned as the
final future.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Now that snapshot orchestration in snapshot_on_all_shards
doesn't use snapshot_manager, get rid of the data structure.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Now that snapshot orchestration is done solely
in snapshot_on_all_shards, the per-shard
snapshot function can be deleted.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Call take_snapshot on each shard and collect the
returns snapshot_file_set.
When all are done, move the vector<snapshot_file_set>
to finalize_snapshot.
All that without resorting to using the snapshot_manager
nor calling table::snapshot.
Both will deleted in the following patches.
Fixes#11132
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Now that seal_snapshot doesn't need to lookup
the snapshot_manager in pending_snapshots to
get to the file_sets, erasing the snapshot_manager
object can be done in table::snapshot which
also inserted it there.
This will make it easier to get rid of it in a later patch.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
and pass it to seal_snapshot, so that the latter won't
need to lookup and access the snapshot_manager object.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Write schema.cql and the files manifest in finalize_snapshot.
Currently call it from table::snapshot, but it will
be called in a later patch by snapshot_on_all_shards.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
To simplify processing of the per-shard file names
for generating the manifest.
We only need to print them to the manifest at the
end of the process, so there's no point in copying
them around in the process, just move the
foreign unique unordered_set.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Currently copying the sstable file names are created
and destroyed on each shard and are copied by the
"coordinator" shards using submit_to, while the
coroutine holds the source on its stack frame.
To prepare for the next patches that refactor this
code so that the coordinator shard will submit_to
each shard to perform `take_snapshot` and return
the set of sstrings in the future result, we need
to wrap the result in a foreign_ptr so it gets
freed on the shard that created it.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
There could be thousands of sstables so we better
cosider yielding in the tight loop that copies
the sstable names into the unordered_set we return.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
