Features
- #24826 Introduces the node config
crash_loop_sleep_sec, which sets the time the broker sleeps before terminating the process when the limit on the number of consecutive times a broker can crash has been reached. This is most useful in Kubernetes environments where setting this value allows customers to have ssh access into a crash looping pod for a short window of time. by @pgellert in #24846
Bug Fixes
- Avoid large allocations for the kafka response sequencing map. by @pgellert in #24742
- Fixes a bug where failing to audit an authentication event could lead to a broker crash. by @pgellert in #24739
- Fixes an issue where transactions incorrectly timeout due incorrect cleanup of evicted producers. by @bharathv in #24878
- Remove partial kvstore snapshots at startup. by @ztlpn in #24843
- #24915 Fixes integer overflow issues when given a schema via the
POST /subject/{subject}/versionwhere version was > INT_MAX or a negative value was provided. by @michael-redpanda in #24917
Improvements
- #24668 stable leadership under load by @mmaslankaprv in #24708
Full Changelog: v24.2.15...v24.2.16