On the read path, the compacting reader is applied only to the sstable
reader. This can cause an expired tombstone from an sstable to be purged
from the request before it has a chance to merge with deleted data in
the memtable leading to data resurrection.
Fix this by checking the memtables before deciding to purge tombstones
from the request on the read path. A tombstone will not be purged if a
key exists in any of the table's memtables with a minimum live timestamp
that is lower than the maximum purgeable timestamp.
Fixes#20916
`perf-simple-query` stats before and after this fix :
`build/Dev/scylla perf-simple-query --smp=1 --flush` :
```
// Before this Fix
// ---------------
94941.79 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59393 insns/op, 24029 cycles/op, 0 errors)
97551.14 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59376 insns/op, 23966 cycles/op, 0 errors)
96599.92 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59367 insns/op, 23998 cycles/op, 0 errors)
97774.91 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59370 insns/op, 23968 cycles/op, 0 errors)
97796.13 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59368 insns/op, 23947 cycles/op, 0 errors)
throughput: mean=96932.78 standard-deviation=1215.71 median=97551.14 median-absolute-deviation=842.13 maximum=97796.13 minimum=94941.79
instructions_per_op: mean=59374.78 standard-deviation=10.78 median=59369.59 median-absolute-deviation=6.36 maximum=59393.12 minimum=59367.02
cpu_cycles_per_op: mean=23981.67 standard-deviation=32.29 median=23967.76 median-absolute-deviation=16.33 maximum=24029.38 minimum=23947.19
// After this Fix
// --------------
95313.53 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59392 insns/op, 24058 cycles/op, 0 errors)
97311.48 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59375 insns/op, 24005 cycles/op, 0 errors)
98043.10 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59381 insns/op, 23941 cycles/op, 0 errors)
96750.31 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59396 insns/op, 24025 cycles/op, 0 errors)
93381.21 tps ( 71.1 allocs/op, 0.0 logallocs/op, 14.1 tasks/op, 59390 insns/op, 24097 cycles/op, 0 errors)
throughput: mean=96159.93 standard-deviation=1847.88 median=96750.31 median-absolute-deviation=1151.55 maximum=98043.10 minimum=93381.21
instructions_per_op: mean=59386.60 standard-deviation=8.78 median=59389.55 median-absolute-deviation=6.02 maximum=59396.40 minimum=59374.73
cpu_cycles_per_op: mean=24025.13 standard-deviation=58.39 median=24025.17 median-absolute-deviation=32.67 maximum=24096.66 minimum=23941.22
```
This PR fixes a regression introduced in ce96b472d3 and should be backported to older versions.
Closesscylladb/scylladb#20985
* github.com:scylladb/scylladb:
topology-custom: add test to verify tombstone gc in read path
replica/table: check memtable before discarding tombstone during read
compaction_group: track maximum timestamp across all sstables
(cherry picked from commit 519e167611)
Backported from #20985 to 6.0
Signed-off-by: Lakshmi Narayanan Sreethar <lakshmi.sreethar@scylladb.com>
Closesscylladb/scylladb#21249
Until we automatically support rebuild for tablets-enabled
keyspaces, warn the user about them.
The reason this is not an error, is that after
increasing RF in a new datacenter, the current procedure
is to run `nodetool rebuild` on all nodes in that dc
to rebuild the new vnode replicas.
This is not required for tablets, since the additional
replicas are rebuilt automatically as part of ALTER KS.
However, `nodetool rebuild` is also run after local
data loss (e.g. due to corruption and removal of sstables).
In this case, rebuild is not supported for tablets-enabled
keyspaces, as tablet replicas that had lost data may have
already been migrated to other nodes, and rebuilding the
requested node will not know about it.
It is advised to repair all nodes in the datacenter instead.
Refs scylladb/scylladb#17575
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
(cherry picked from commit ed1e9a1543)
Closesscylladb/scylladb#20724
During split prepare phase, there will be more than 1 compaction group with
overlapping token range for a given replica.
Assume tablet 1 has sstable A containing deleted data, and sstable B containing
a tombstone that shadows data in A.
Then split starts:
sstable B is split first, and moved from main (unsplit) group to a
split-ready group
now compaction runs in split-ready group before sstable A is split
tombstone GC logic today only looks at underlying group, so compaction is step
2 will discard the deleted data in A, since it belongs to another group (the
unsplit one), and so the tombstone can be purged incorrectly.
To fix it, compaction will now work with all uncompacting sstables that belong
to the same replica, since tombstone GC requires all sstables that possibly
contain shadowed data to be available for correct decision to be made.
Fixes https://github.com/scylladb/scylladb/issues/20044.
Please replace this line with justification for the backport/* labels added to this PR
Branches 6.0, 6.1 and 6.2 are vulnerable, so backport is needed.
(cherry picked from commit bcd358595f)
(cherry picked from commit 93815e0649)
Refs https://github.com/scylladb/scylladb/pull/20939Closesscylladb/scylladb#21204
* github.com:scylladb/scylladb:
replica: Fix tombstone GC during tablet split preparation
service: Improve error handling for split
During split prepare phase, there will be more than 1 compaction group with
overlapping token range for a given replica.
Assume tablet 1 has sstable A containing deleted data, and sstable B containing
a tombstone that shadows data in A.
Then split starts:
1) sstable B is split first, and moved from main (unsplit) group to a
split-ready group
2) now compaction runs in split-ready group before sstable A is split
tombstone GC logic today only looks at underlying group, so compaction is step
2 will discard the deleted data in A, since it belongs to another group (the
unsplit one), and so the tombstone can be purged incorrectly.
To fix it, compaction will now work with all uncompacting sstables that belong
to the same replica, since tombstone GC requires all sstables that possibly
contain shadowed data to be available for correct decision to be made.
Fixes#20044.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
(cherry picked from commit 93815e0649)
Currently, a pending replica that applies a write on a table that has
materialized views, will build all the view updates as a normal replica,
only to realize at a late point, in db::view::get_view_natural_endpoint(),
that it doesn't have a paired view replica to send the updates to. It will
then either drop the view updates, or send them to a pending view
replica, if such exists.
This work is unnecessary since it may be dropped, and even if there is a
pending view replica to send the updates to, the updates that are built
by the pending replica may be wrong since it may have incomplete
information.
This commit fixes the inefficiency by skipping the view update building
step when applying an update on a pending replica.
The metric total_view_updates_on_wrong_node is added to count the cases
that a view update is determined to be unnecessary.
The test reproduces the scenario of writing to a table and applying
the update on a pending replica, and verifies that the pending replica
doesn't try to build view updates.
Fixesscylladb/scylladb#19152Closesscylladb/scylladb#19488Fixesscylladb/scylladb#20787
(cherry picked from commit 08b29460fc)
Closesscylladb/scylladb#20934
Consider the following:
```
T
0 split prepare starts
1 repair starts
2 split prepare finishes
3 repair adds unsplit sstables
4 repair ends
5 split executes
```
If repair produces sstable after split prepare phase, the replica will not split that sstable later, as prepare phase is considered completed already. That causes split execution to fail as replicas weren't really prepared. This also can be triggered with load-and-stream which shares the same write (consumer) path.
The approach to fix this is the same employed to prevent a race between split and migration. If migration happens during prepare phase, it can happen source misses the split request, but the tablet will still be split on the destination (if needed). Similarly, the repair writer becomes responsible for splitting the data if underlying table is in split mode. That's implemented in replica::table for correctness, so if node crashes, the new sstable missing split is still split before added to the set.
Fixes https://github.com/scylladb/scylladb/issues/19378.
Fixes https://github.com/scylladb/scylladb/issues/19416.
Please replace this line with justification for the backport/* labels added to this PR
(cherry picked from commit 239344ab55)
(cherry picked from commit 74612ad358)
Refs https://github.com/scylladb/scylladb/pull/19427Closesscylladb/scylladb#20593
* github.com:scylladb/scylladb:
tablets: Fix race between repair and split
compaction: Allow "offline" sstable to be split
Consider the following:
T
0 split prepare starts
1 repair starts
2 split prepare finishes
3 repair adds unsplit sstables
4 repair ends
5 split executes
If repair produces sstable after split prepare phase, the replica
will not split that sstable later, as prepare phase is considered
completed already. That causes split execution to fail as replicas
weren't really prepared. This also can be triggered with
load-and-stream which shares the same write (consumer) path.
The approach to fix this is the same employed to prevent a race
between split and migration. If migration happens during prepare
phase, it can happen source misses the split request, but the
tablet will still be split on the destination (if needed).
Similarly, the repair writer becomes responsible for splitting
the data if underlying table is in split mode. That's implemented
in replica::table for correctness, so if node crashes, the new
sstable missing split is still split before added to the set.
Fixes#19378.
Fixes#19416.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
(cherry picked from commit 74612ad358)
Currently, database::tables_metadata::add_table needs to hold a write
lock before adding a table. So, if we update other classes keeping
track of tables before calling add_table, and the method yields,
table's metadata will be inconsistent.
Set all table-related info in tables_metadata::add_table_helper (called
by add_table) so that the operation is atomic.
Analogically for remove_table.
Fixes: #19833.
(cherry picked from commit 483d89ed6d)
Closesscylladb/scylladb#20245
It was added to make integration of storage groups easier, but it's
complicated since it's another source of truth and we could have
problems if it becomes inconsistent with the group map.
Fixes#18506.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
(cherry picked from commit ad5c5bca5f)
compaction_group sits in replica layer and compaction layer is
supposed to talk to it through compaction::table_state only.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
(cherry picked from commit b8bd4c51c2)
Retrieval of tablet stats must be serialized with mutation to token metadata, as the former requires tablet id stability.
If tablet split is finalized while retrieving stats, the saved erm, used by all shards, can have a lower tablet count than the one in a particular shard, causing an abort as tablet map requires that any id feeded into it is lower than its current tablet count.
Fixes https://github.com/scylladb/scylladb/issues/18085.
(cherry picked from commit abcc68dbe7)
(cherry picked from commit 551bf9dd58)
(cherry picked from commit e7246751b6)
Refs https://github.com/scylladb/scylladb/pull/18287Closesscylladb/scylladb#19095
* github.com:scylladb/scylladb:
topology_experimental_raft/test_tablets: restore usage of check_with_down
test: Fix flakiness in topology_experimental_raft/test_tablets
service: Use tablet read selector to determine which replica to account table stats
storage_service: Fix race between tablet split and stats retrieval
tablet snapshot, used by migration, can race with compaction and
can find files deleted. That won't cause data loss because the
error is propagated back into the coordinator that decides to
retry streaming stage. So the consequence is delayed migration,
which might in turn reduce node operation throughput (e.g.
when decommissioning a node). It should be rare though, so
shouldn't have drastic consequences.
Fixes#18977.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
(cherry picked from commit b396b05e20)
Closesscylladb/scylladb#19008
If tablet split is finalized while retrieving stats, the saved erm, used by all
shards, will be invalidated. It can either cause incorrect behavior or
crash if id is not available.
It's worked by feeding local tablet map into the "coordinator"
collecting stats from all shards. We will also no longer have a snapshot
of erm shared between shards to help intra-node migration. This is
simplified by serializing token metadata changes and the retrieval of
the stats (latter should complete pretty fast, so it shouldn't block
the former for any significant time).
Fixes#18085.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
(cherry picked from commit abcc68dbe7)
Currently, we do not explicitly set a scheduling group for the schema
commitlog which causes it to run in the default scheduling group (called
"main"). However:
- It is important and significant enough that it should run in a
scheduling group that is separate from the main one,
- It should not run in the existing "commitlog" group as user writes may
sometimes need to wait for schema commitlog writes (e.g. read barrier
done to learn the schema necessary to interpret the user write) and we
want to avoid priority inversion issues.
Therefore, introduce a new scheduling group dedicated to the schema
commitlog.
Fixes: scylladb/scylladb#15566Closesscylladb/scylladb#18715
There's a set of API endpoints that toggle per-table auto-compaction and tombstone-gc booleans. They all live in two different .cc files under api/ directory and duplicate code of each other. This PR generalizes those handlers, places them next to each other, fixes leak on stop and, as a nice side effect, enlightens database.hh header.
Closesscylladb/scylladb#18703
* github.com:scylladb/scylladb:
api,database: Move auto-compaction toggle guard
api: Move some table manipulation helpers from storage_service
api: Move table-related calls from storage_service domain
api: Reimplement some endpoints using existing helpers
api: Lost unset of tombstone-gc endpoints
Toggling per-table auto-compaction enabling bit is guarded with
on-database boolean and raii guard. It's only used by a single
api/column_family.cc file, so it can live there.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
This series is a reupload of #13792 with a few modifications, namely a test is added and the conflicts with recent tablet related changes are fixed.
See https://github.com/scylladb/scylladb/issues/12379 and https://github.com/scylladb/scylladb/pull/13583 for a detailed description of the problem and discussions.
This PR aims to extend the existing throttling mechanism to work with requests that internally generate a large amount of view updates, as suggested by @nyh.
The existing mechanism works in the following way:
* Client sends a request, we generate the view updates corresponding to the request and spawn background tasks which will send these updates to remote nodes
* Each background task consumes some units from the `view_update_concurrency_semaphore`, but doesn't wait for these units, it's just for tracking
* We keep track of the percent of consumed units on each node, this is called `view update backlog`.
* Before sending a response to the client we sleep for a short amount of time. The amount of time to sleep for is based on the fullness of this `view update backlog`. For a well behaved client with limited concurrency this will limit the amount of incoming requests to a manageable level.
This mechanism doesn't handle large DELETE queries. Deleting a partition is fast for the base table, but it requires us to generate a view update for every single deleted row. The number of deleted rows per single client request can be in the millions. Delaying response to the request doesn't help when a single request can generate millions of updates.
To deal with this we could treat the view update generator just like any other client and force it to wait a bit of time before sending the next batch of updates. The amount of time to wait for is calculated just like in the existing throttling code, it's based on the fullness of `view update backlogs`.
The new algorithm of view update generation looks something like this:
```c++
for(;;) {
auto updates = generate_updates_batch_with_max_100_rows();
co_await seastar::sleep(calculate_sleep_time_from_backlogs());
spawn_background_tasks_for_updates(updates);
}
```
Fixes: https://github.com/scylladb/scylladb/issues/12379Closesscylladb/scylladb#16819
* github.com:scylladb/scylladb:
test: add test for bad_allocs during large mv queries
mv: throttle view update generation for large queries
exceptions: add read_write_timeout_exception, a subclass of request_timeout_exception
db/view: extract view throttling delay calculation to a global function
view_update_generator: add get_storage_proxy()
storage_proxy: make view backlog getters public
With intra-node migration, all the movement is local, so we can make
streaming faster by just cloning the sstable set of leaving replica
and loading it into the pending one.
This cloning is underlying storage specific, but s3 doesn't support
snapshot() yet (th sstables::storage procedure which clone is built
upon). It's only supported by file system, with help of hard links.
A new generation is picked for new cloned sstable, and it will
live in the same directory as the original.
A challenge I bumped into was to understand why table refused to
load the sstable at pending replica, as it considered them foreign.
Later I realized that sharder (for reads) at this stage of migration
will point only to leaving replica. It didn't fail with mutation
based streaming, because the sstable writer considers the shard --
that the sstable was written into -- as its owner, regardless of what
sharder says. That was fixed by mimicking this behavior during
loading at pending.
test:
./test.py --mode=dev intranode --repeat=100 passes.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Require users to specify whether we want shard for reads or for writes
by switching to appropriate non-deprecated variant.
For example, shard_of() can be replaced with shard_for_reads() or
shard_for_writes().
The next_shard/token_for_next_shard APIs have only for-reads variant,
and the act of switching will be a testimony to the fact that the code
is valid for intra-node migration.
The namespace replica is broken in the middle with sstable_list alias,
while the latter can be declared earlier
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Closesscylladb/scylladb#18664
This patch adds a test for reproducing issue #12379, which is
being fixed in #16819.
The test case works by creating a table with a materialized
view, and then performing a partition delete query on it.
At the same time, it uses injections to limit the memory
to a level lower than usual, in order to increase the
consistency of the test, and to limit its runtime.
Before #16819, the test would exceed the limit and fail,
and now the next allocation is throttled using a sleep.
Closesscylladb/scylladb#18616
* github.com:scylladb/scylladb:
replica: Make it explicit table's sstable set is immutable
replica: avoid reallocations in tablet_sstable_set
replica: Avoid compound set if only one sstable set is filled
Currently empty storage_groups are allocated for tablets that are
not on this shard.
Allocate storage groups dynamically, i.e.:
- on table creation allocate only storage groups that are on this
shard;
- allocate a storage group for tablet that is moved to this shard;
- deallocate storage group for tablet that is cleaned up.
Stop compaction group before it's deallocated.
Add a flag to table::cleanup_tablet deciding whether to deallocate
sgs and use it in commitlog tests.
Add rwlock which prevents storage groups from being added/deleted
while some other layers itereates over them (or their compaction
groups).
Add methods to iterate over storage groups with the lock held.
Pass compaction group id to
shard_reshaping_compaction_task_impl::reshape_compaction_group.
Modify table::as_table_state to return table_state of the given
compaction group.
Currently we have a single method -- detach_column_family() -- which
does something with each semaphore. Soon there will be another one.
Introduce a method to do something with all semaphores, to make this
smoother. Enterprise has a different set of semaphores, and this will
reduce friction.
The populate_views() and generate_and_propagate_view_updates() both naturally belong to view_update_generator -- they don't need anything special from table itself, but rather depend on some internals of the v.u.generator itself.
Moving them there lets removing the view concurrency semaphore from keyspace and table, thus reducing the cross-components dependencies.
Closesscylladb/scylladb#18421
* github.com:scylladb/scylladb:
replica: Do not carry view concurrency semaphore pointer around
view: Get concurrency semaphore via database, not table
view_update_generator: Mark mutate_MV() private
view: Move view_update_generator methods' code
view: Move table::generate_and_propagate_view_updates into view code
view: Move table::populate_views() into view_update_generator class
There are two places that get total:live stats for a table snapshot --
database::get_snapshot_details() and table::get_snapshot_details(). Both
do pretty similar thing -- walk the table/snapshots/ directory, then
each of the found sub-directory and accumulate the found files' sizes
into snapshot details structure.
Both try to tell total from live sizes by checking whether an sstable
component found in snapshots is present in the table datadir. The
database code does it in a more correct way -- not just checks the file
presense, but also compares if it's a hardlink on the snapshot file,
while the table code just checks if the file of the same name exists.
This patch does both -- makes both database and table call the same
helper method for a single snapshot details, and makes the generalized
version use more elaborated collision check, thus fixing the per-table
details getting behavior.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Closesscylladb/scylladb#18347
The _view_update_concurrency_sem field on database propagates itself via
keyspace config down to table config and view_update_generator then
grabs one via table:: helper. That's an overkil, view_update_generator
has direct reference on the database and can get this semaphore from
there.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Similarly to populate_views() method, this one also naturally belongs to
view_update_generator class.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
The method in question has little to do with table, effectively it only
needs stats and consurrency semaphore. And the semaphore in question is
obtained from table indirectly, it really resides on database. On the
other hand, the method carries lots of bits from db::view, e.g. the
view_update_builder class, memory_usage_of() helper and a bit more.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
There's a database::get_snapshot_details() method that returns collection of all snapshots for all ks.cf out there and there are several *snapshot_details* aux structures around it. This PR keeps only one "details" and cleans up the way it propagates from database up to the respective API calls.
Closesscylladb/scylladb#18317
* github.com:scylladb/scylladb:
snapshot_ctl: Brush up true_snapshots_size() internals
snapshot_ctl: Remove unused details struct
snapshot_ctl: No double recoding of details
database,snapshots: Move database::snapshot_details into snapshot_ctl
database,snapshots: Make database::get_snapshot_details() return map, not vector
table,snapshots: Move table::snapshot_details into snapshot_ctl
So that it's in-sync with table::get_snapshot_details(). Next patches
will improve this place even further.
Also, there can be many snapshots and vector can grow large, but that's
less of an issue here.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Now database is constructed with sharded<directory_semaphore>, but it no
longer needs sharded, local is enough.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
It was only used by table taking snapshot code. Now it uses sstables
manager's reference and database can stop carrying it around.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
storage_group_id_for_token() was only needed from within
tablet_storage_group_manager, so we can kill
table::storage_group_id_for_token().
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closesscylladb/scylladb#18134
Make a specialized sstable_set for tablets
via tablet_storage_group_manager::make_sstable_set.
This sstable set takes a snapshot of the storage_groups
(compound) sstable_sets and maps the selected tokens
directly into the tablet compound_sstable_set.
This sstable_set provides much more efficient access
to the table's sstable sets as it takes advantage of the disjointness
of sstable sets between tablets/storage_groups, and making it is cheaper
that rebuilding a complete partitioned_sstable_set from all sstables in the table.
Fixes#16876
Cassandra-stress setup:
```
$ sudo cpupower frequency-set -g userspace
$ build/release/scylla (developer-mode options) --smp=16 --memory=8G --experimental-features=consistent-topology-changes --experimental-features=tablets
cqlsh> CREATE KEYSPACE keyspace1 WITH replication={'class':'NetworkTopologyStrategy', 'replication_factor':1} AND tablets={'initial':2048};
$ ./tools/java/tools/bin/cassandra-stress write no-warmup n=10000000 -pop 'seq=1...10000000' -rate threads=128
$ scylla-api-client system drop_sstable_caches POST
$ ./tools/java/tools/bin/cassandra-stress read no-warmup duration=60s -pop 'dist=uniform(1..10000000)' -rate threads=128
$ scylla-api-client system drop_sstable_caches POST
$ ./tools/java/tools/bin/cassandra-stress mixed no-warmup duration=60s -pop 'dist=uniform(1..10000000)' -rate threads=128
```
Baseline (0a7854ea4d) vs. fix (0c2c00f01b)
Throughput (op/s):
workload | baseline | fix
---------|----------|----------
write | 76,806 | 100,787
read | 34,330 | 106,099
mixed | 32,195 | 79,246
Closesscylladb/scylladb#17149
* github.com:scylladb/scylladb:
table: tablet_storage_group_manager: make tablet_sstable_set
storage_group_manager: add make_sstable_set
tablet_storage_group_manager: handle_tablet_split_completion: pre-calc new_tablet_count
table: tablet_storage_group_manager: storage_group_of: do not validate in release build mode
table: move compaction_group_list and storage_group_vector to storage_group_manager
compaction_group::table_state: get_group_id: become self-sufficient
compaction_group, table: make_compound_sstable_set: declare as const
tablet_storage_group_manager: precalculate my_host_id and _tablet_map
table: coroutinize update_effective_replication_map
Make a specialized sstable_set for tablets
via tablet_storage_group_manager::make_sstable_set.
This sstable set takes a snapshot of the storage_groups
(compound) sstable_sets and maps the selected tokens
directly into the tablet compound_sstable_set.
Refs #16876
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
