What yugabyte/yugabyte-db shipped
Written by FoxPlug from public releases; not affiliated with Yugabyte. An automatic summary of the public release, pull request and commit data of github.com/yugabyte/yugabyte-db. Yugabyte did not write it and does not use or endorse FoxPlug. Every line links to the public change it describes.
Get a weekly update like this for your product, free
Week of September 21, 2026
What shipped
- YSQL catalog cache preloading now uses a 10 MiB aggregate response budget instead of 1,024-row page limits, reducing serial master round trips after DDL on large catalogs in multi-region clusters. Pull request #34178
- DocDB now supports an optional
max_num_replicasconstraint on placement blocks for cluster, table, and tablespace policies to control replica distribution. Pull request #32828 - DocDB thin client now supports UINT32 columns, enabling it to open tables keyed by oids. Pull request #34152
- LongOperationTracker lock contention reduced by replacing a single process-wide mutex with per-entry locking. Pull request #33273
- DocDB now reports per-drive write throughput alongside fsync latency metrics for better device performance interpretation. Pull request #34082
- YSQL fixed a NULL crash in row comparison binds when comparing rows with NULL varlena columns. Release
- DocDB thin client stack trace collection now uses libunwind to prevent deadlocks during creation on certain allocators. Pull request #34198
- YSQL webserver test isolation on macOS fixed by reordering pkill options to prevent unrelated process kills during parallel test runs. Pull request #33588
- YBA now allows skipping storage configuration validation via runtime config for backup storage that the YBA host cannot directly reach. Pull request #34081
- Build system now reruns new tests on integration lanes to catch flakiness introduced at commit time. Pull request #34029
Why it matters
This week ships improvements to query performance and reliability: catalog preloading is faster on large clusters, placement constraints provide better control over replica distribution, and thin client support for more column types enables new use cases. Several fixes address crashes, deadlocks, and test isolation issues that improve stability.
Changelog entry
- YSQL: Fixed PgSchemaVersionMismatchBackfillTest.BackfillSurfaceAsSerializationFailure Release
- YSQL: Fixed NULL crash in row comparison binds with varlena key columns Release
- YSQL: Fixed TestWebserverKill killing unrelated processes on macOS Pull request #33588
- YSQL: Batch catalog preload requests by 10 MiB byte limits instead of 1,024-row limits Pull request #34178
- DocDB: Added UINT32 column support to
yb_thin_clientfor oid-keyed tables Pull request #34159 - DocDB: Colocated tombstone cache now persists across schema GC and backfill Pull request #34153
- DocDB: Non-ASCII bytes now properly escaped in Slice::ToDebugString regardless of locale Pull request #33568
- DocDB: Thin client stack traces now collected through libunwind to prevent deadlocks Pull request #34198
- YSQL: Fixed PgSchemaVersionMismatchBackfillTest.BackfillSurfacesAsSerializationFailure under TSAN Pull request #34206
- DocDB: Raft catch-up unstuck by advancing
last_received_current_leaderon fully-deduplicated requests Pull request #34085 - DocDB: Sort indicator now visible on table headers in web UI Pull request #34129
- DocDB: Per-drive write throughput and fsync latency metrics added Pull request #34082
- DocDB: Placement blocks now support optional
max_num_replicasconstraint Pull request #32828 - DocDB: LongOperationTracker lock contention reduced with per-entry locking Pull request #33273
- Build: macOS builds migrated from macOS 14 to macOS 15 Pull request #34053
New in yugabyte-db: faster DDL on large catalogs with adaptive response budgets, placement block replica limits, thin client UINT32 support, reduced lock contention in operation tracking, and per-drive fsync metrics.
This week's yugabyte-db updates focus on performance and reliability improvements. Catalog preloading now adapts to response size budgets instead of fixed row limits, reducing latency after DDL on large clusters. New placement block constraints give operators fine-grained control over replica distribution. The thin client gains support for UINT32 columns to open tables keyed by oids. Lock contention improvements reduce blocking in operation tracking. Per-drive write throughput metrics alongside fsync latency provide clearer device performance visibility.
Week of September 14, 2026
What shipped
- Distributed trace context now propagates across RPC boundaries, with the request header carrying an optional trace-context field that flows from caller to callee. Pull request #33112
- The
dist_tracelibrary was reworked into a span-and-scope API with entry points for starting traces and re-establishing parent context. Pull request #33116 - Trace context now carries across thread boundaries so work handed to worker, reactor, scheduler, or Raft apply threads continues the originating trace. Pull request #33114
- Trace context propagates over the PG-tserver shared memory exchange so requests bypassing the RPC stack still join the caller's trace. Pull request #33113
- The traceparent header now propagates to PG backends via the startup packet so backend initialization and statements root under the originating trace. Pull request #33115
- Per-tablet in-memory aggregation was added for SST statistics to answer questions across multiple SSTfiles rather than single-file records. Pull request #33887
- Index backfill failures now report with PostgreSQL SQLSTATE codes instead of the generic XX000 error code. Pull request #33748
- CheckpointMultipleDatabases in xCluster replication was capped at 64 databases to prevent excessive concurrent DDL operations. Pull request #33933
- Python 3.14 was added to the supported interpreter versions for YBA with updated dependency pins. Pull request #33993
- YBA now fails fast with a clear error when the master leader is missing instead of silently dereferencing null values. Pull request #33840
Why it matters
Distributed tracing infrastructure across RPC boundaries and thread hops enables operators to track requests through the entire system. Improvements to error reporting, resource limits, and safety checks make the database more reliable and easier to debug.
Changelog entry
- Distributed trace context now propagates across RPC boundaries Pull request #33112
- Reworked
dist_tracelibrary into span-and-scope API Pull request #33116 - Trace context carries across thread boundaries Pull request #33114
- Trace context propagates over PG-tserver shared memory exchange Pull request #33113
- Traceparent propagates to PG backends via startup packet Pull request #33115
- Per-tablet in-memory aggregation for SST statistics Pull request #33887
- Index backfill failures report with PostgreSQL SQLSTATE codes Pull request #33748
- CheckpointMultipleDatabases capped at 64 databases in xCluster Pull request #33933
- Python 3.14 added to YBA supported versions Pull request #33993
- YBA fails fast with clear error when master leader is missing Pull request #33840
Distributed tracing now flows across RPC boundaries, thread hops, and shared memory exchanges. Better error codes and safety checks improve debugging and prevent silent failures.
This week's work strengthens distributed tracing throughout YugabyteDB. Trace context now propagates across RPC boundaries, thread boundaries, and the PG-tserver shared memory path—enabling operators to follow requests through the entire system. We also improved error reporting with proper PostgreSQL error codes, added resource limits for xCluster operations, and added safety checks to fail fast when critical components are missing.