[RELEASE] ScyllaDB 6.2.3

Release Notes
, ,
1

The ScyllaDB team announces ScyllaDB Open Source 6.2.3, a bugfix patch release of the ScyllaDB 6.2 stable branch. ScyllaDB Open Source 6.2.3, like all past and future 6.x.y releases, is backward compatible and supports rolling upgrades.

Users are encouraged to upgrade to 6.2.3.

Related links:

Issue fixed in this release:

Stability

  • boost/secondary_index_test failed with 0x28be8a4 #20772. The root cause is an init race condition between view builder and group 0 builder. The first depends on the second.

Related issues: #21534

  • Get_service_levels is an internal function that can be called by internal services, and directly by the user using LIST SERVICE LEVELS. The fix increases the timeout for the first (internal), where response time is less important. #20483
  • Abort in service::storage_service::handle_state_normal in test_decommission_after_changing_node_ip. This is a leftover from moving ScyllaDB to use host-id instead of IPs everywhere. #21930
  • Error raft_topology - topology change coordinator fiber got error(Cannot achieve consistency level for cl ALL. Requires 3, alive 2) during upgrade latest node in cluster #20754. The root cause is in the view builder.
  • malformed_sstable_exception due to reclaim and reload of bloom filters from unlinked sstables #21887
  • The gossip code will now clean up nodes that died before they could join Raft. #20082
  • repair: task_manager_module::is_aborted accesses task map of a caller #22156
  • repair: tablet_repair_task_impl keeps erm for task_ttl seconds #21503
  • Service level cache may not be populated at startup #21763
  • test_topology_recovery_basic fails when validation queries are done with all_pages=True #19101
  • MV: test_view_build_status_with_replace is flaky. Root cause is race condition with very small (short) MV build #20638
  • transport/server: async callback is unsafe in for_each_gently() #21801. Revert a change in 6.2.0

Tooling

  • An sstable scrub results in rewriting sstables, but did not remove the original sstable from metadata, causing a later reshape compaction to fail. This is now fixed. #20030

Logs

  • raft: improve logs for abort while waiting for apply #22159

Performance

  • ScyllaDB tracks whether read requests are waiting for CPU or I/O. In one case, a disk read from the primary index was considered to be waiting on CPU, which reduced concurrency. This is now fixed. #21325