mirrors/scylladb
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
ec42fdfd01ae207fac89b9929704bf4571b8d99a
scylladb/service
History
Ferenc Szili f70ca9a406 load_stats: implement the optimized sum of tablet sizes 
PR #28703 was merged into master but not with the latest version of the
changes. This patch is an incremental fix for this.

Currently, the elements of the tablet_sizes_per_shard vector are
incremented in separate shards. This is prone to false sharing of cache
lines, and ping-pong of memory, which leads to reduced performance.

In this patch, in order to avoid cache line collisions while updating
the sum of tablet sizes per shard, we align the counter to 64 bytes.

Fixes: SCYLLADB-678

Closes scylladb/scylladb#28757
2026-02-24 22:19:25 +01:00
..
broadcast_tables/experimental
treewide: Move replica related files to replica directory
2025-09-18 08:00:35 +03:00
direct_failure_detector
treewide: #include Seastar headers with angle brackets
2026-01-13 14:56:15 +02:00
pager
mv: allow setting concurrency in PRUNE MATERIALIZED VIEW
2025-11-27 00:02:28 +01:00
paxos
paxos_state: get_replica_lock: remove shard check
2025-10-31 21:37:39 +01:00
qos
Populate all sl:* groups into dedicated top-level supergroup
2026-01-21 14:14:48 +02:00
raft
raft topology: make some assertions non-crashing
2026-02-12 13:10:03 +01:00
strong_consistency
strong consistency: implement coordinator::query()
2026-01-21 14:56:01 +01:00
address_map.hh
code: Stop using seastar::compat::source_location
2025-11-27 19:10:11 +02:00
cache_hitrate_calculator.hh
…
cas_shard.hh
…
client_routes.cc
db: api: service: Fix ClientConnectorError in test_client_routes The bug was caused by capturing local variables by reference in lambdas passed to with_retry(), which is a coroutine. When the coroutine suspends, the lambda frame exits and the referenced locals are destroyed, leading to use-after-lifetime issues. This change fixes the problem by ensuring safe ownership across suspension points and also refactors how route_keys and route_entries are passed from the caller. Previously they were passed as const lvalue references, which cannot be moved and therefore ended up being repeatedly copied across function calls and lambda invocations. The new approach avoids unnecessary copies and makes the lifetime semantics explicit and safe.
2025-12-22 14:52:47 +02:00
client_routes.hh
db: api: service: Fix ClientConnectorError in test_client_routes The bug was caused by capturing local variables by reference in lambdas passed to with_retry(), which is a coroutine. When the coroutine suspends, the lambda frame exits and the referenced locals are destroyed, leading to use-after-lifetime issues. This change fixes the problem by ensuring safe ownership across suspension points and also refactors how route_keys and route_entries are passed from the caller. Previously they were passed as const lvalue references, which cannot be moved and therefore ended up being repeatedly copied across function calls and lambda invocations. The new approach avoids unnecessary copies and makes the lifetime semantics explicit and safe.
2025-12-22 14:52:47 +02:00
client_state.cc
auth: add CDC streams and timestamps to vector search permissions
2026-02-04 09:10:08 +01:00
client_state.hh
service: remove unused has_schema_access
2026-01-28 10:18:26 +02:00
CMakeLists.txt
strong_consistency: add coordinator
2026-01-21 14:56:01 +01:00
endpoint_lifecycle_subscriber.hh
service: transport: add CLIENT_ROUTES_CHANGE event
2025-12-15 18:19:37 +01:00
load_broadcaster.hh
code: Replace distributed<> with sharded<>
2025-09-19 12:22:51 +02:00
load_meter.hh
code: Replace distributed<> with sharded<>
2025-09-19 12:22:51 +02:00
maintenance_mode.hh
…
mapreduce_service.cc
treewide: Move query related files to a new query directory
2025-09-16 23:40:47 +03:00
mapreduce_service.hh
code: Replace distributed<> with sharded<>
2025-09-19 12:22:51 +02:00
memory_limiter.hh
…
migration_listener.hh
migration_listener: fix deadlock in nested notifications
2025-12-17 14:00:28 +01:00
migration_manager.cc
gossiper: Export its scheduling group for those who need it
2026-01-28 18:29:33 +03:00
migration_manager.hh
migration_manager: Reorder members
2026-01-28 18:29:33 +03:00
misc_services.cc
replica/table: keep track of total pre-compression file size
2025-11-13 00:49:57 +01:00
query_state.hh
…
session.cc
…
session.hh
…
state_id.hh
…
storage_proxy_fwd.hh
…
storage_proxy_stats.hh
storage_proxy_stats: add fenced_out_requests metric
2025-09-15 11:24:53 +02:00
storage_proxy.cc
storage_proxy: hold erms in replica handlers
2026-02-16 08:57:42 +01:00
storage_proxy.hh
hints: Provide explicit scheduling group for hint_sender
2026-01-27 12:50:11 +02:00
storage_service.cc
load_stats: implement the optimized sum of tablet sizes
2026-02-24 22:19:25 +01:00
storage_service.hh
topology_coordinator: pass raft_topology_cmd by value
2026-02-16 08:57:42 +01:00
tablet_allocator_fwd.hh
…
tablet_allocator.cc
Merge 'tablets: Reduce per-shard migration concurrency to 2' from Tomasz Grabiec
2026-02-19 15:31:43 +02:00
tablet_allocator.hh
Merge 'Improve load balancer logging and other minor cleanups' from Tomasz Grabiec
2026-01-29 08:25:17 +02:00
tablet_operation.hh
…
task_manager_module.cc
service: tasks: fix type of global_topology_request_virtual_task
2026-01-16 11:36:21 +01:00
task_manager_module.hh
tasks: service: add global_topology_request_virtual_task
2025-12-16 13:31:22 +01:00
topology_coordinator.cc
Merge 'raft ropology: prevent crashes of multiple nodes' from Patryk Jędrzejczak
2026-02-19 16:50:03 +01:00
topology_coordinator.hh
topology_coordinator: pass raft_topology_cmd by value
2026-02-16 08:57:42 +01:00
topology_guard.hh
…
topology_mutation.cc
raft topology: make some assertions non-crashing
2026-02-12 13:10:03 +01:00
topology_mutation.hh
db: service: add paused_rf_change_requests to system.topology
2025-12-16 13:25:38 +01:00
topology_state_machine.cc
topology: Protect against empty cancelation reason
2026-01-18 15:36:05 +01:00
topology_state_machine.hh
tasks, topology: Make pending node operations abortable
2026-01-18 15:36:05 +01:00
view_update_backlog_broker.hh
…