[RELEASE] ScyllaDB 2025.4.7

Release Notes
, ,
1

The ScyllaDB team is pleased to announce the release of ScyllaDB 2025.4.7, a production-ready patch release for ScyllaDB 2025.4 Feature Release.

Note there is a new Long Term Support (LTS) release 2026.1. You are welcome to upgrade to it for the latest and greatest features.

Related Links

Bug Fixes

Materialized Views

  • An issue was identified in view_update_generator::mutate_MV where a keyspace reference was incorrectly held across a preemption point. This improper handling could cause incorrect results when updating Materialized Views, which was previously detected by an AddressSanitizer use-after-poison error. The fix ensures that the keyspace reference is not held during preemption, resolving the fundamental correctness problem and significantly enhancing the stability of Materialized View updates.
    scylladb#28925, scylladb#29095

Monitoring

  • Metrics used to track garbage-collected (GC) tombstones were inaccurate, as they incorrectly accounted for all processed tombstones rather than only those that had expired. The fix updates the tombstone GC metrics to only report on expired tombstones, providing administrators with accurate visibility into compaction effectiveness.
    scylladb#28743

Raft

  • A node could experience a hang during the join process for Raft group 0 if it attempted to join the group before the post_server_start phase was fully complete. The updated join logic resolves this issue, ensuring a more reliable and smoother node bootstrapping experience.
    scylladb#29308

  • Stale snapshot transfers in Raft-managed clusters were not being aborted when the Raft term changed, leading to wasted resources. The system now immediately aborts these transfers upon a Raft term change, improving cluster stability and efficiency by ensuring Raft groups operate only on current data.
    scylladb#29357

Service Levels

  • When migrating service levels, the v2 migration process would not correctly complete if the underlying legacy table was empty. This issue has been fixed by correctly marking the v2 migration as complete on empty legacy tables, which streamlines the migration process and prevents unexpected delays during upgrades.
    scylladb#29352