Redpanda Files

Bug Fixes

• Fixed a cloud-topics epoch fencing race where an epoch bump whose replication failed could later admit a stale-epoch write, poisoning the partition log and crashing all replicas of the partition. by @Lazin in #31140
• Fixes a bug in OffsetForLeaderEpoch handling with tiered topics which could cause consumers to reset their fetch position to the beginning of the log held in cloud storage. by @WillemKauf in #31074
• Fixes a potential crash in DescribeLogDirs by @WillemKauf in #31196
• HTTP Proxy: Consumer group fetches no longer fail indefinitely with offset_out_of_range after retention moves a topic's log start offset past 0; the consumer now recovers to the earliest available offset (auto.offset.reset=earliest). by @bartoszpiekny-redpanda in #31207
• Lifecycle fix on groups by @joe-redpanda in #31182
• #30885 Fixed a raft pre-vote livelock where a follower recovering behind a live leader could repeatedly start elections, leaving partitions under-replicated and stalling acks=all produce with request timeouts. by @stevenpall in #30886
• #31238 Fixes a bug in which internal topics could be auto-created with the wrong topic properties by @WillemKauf in #31239
• fix allocation double free on no-op rf changes by @joe-redpanda in #31116

Improvements

• Added a `schema_registry_replay_on_startup` cluster property (default off) that hydrates the Schema Registry store at broker start-up rather than lazily on the first request. by @sjust-redpanda in #31174
• Avoid oversized allocations for some of the `schema_registry` and `pandaproxy` handler interfaces. by @WillemKauf in #31200
• HTTP Proxy: a consumer group now resumes from its committed offset on a fresh consumer instance instead of re-reading from the earliest available offset. by @bartoszpiekny-redpanda in #31225
• Optimize the `schema_registry` recovery path by caching references when canonicalizing schemas by @WillemKauf in #31193
• Schema Registry now replays the internal `_schemas` topic exactly once when recovering on startup, instead of running redundant concurrent replays; cold start of a large registry is significantly faster. by @sjust-redpanda in #31174

Full Changelog: https://github.com/redpanda-data/redpanda/compare/v26.1.13...v26.1.14