[RELEASE] ScyllaDB 2025.1.11

Release Notes
, ,
1

The ScyllaDB team announces ScyllaDB 2025.1.11, a bug-fix production-ready patch release for ScyllaDB 2025.1 LTS Release.

Note there is a new Short-Term Support (STS) Feature release 2025.4. You are welcome to upgrade to it for the latest and greatest features, or stay on the 2025.1 track for long-term support.

Related Links

Bug Fixes

API

  • Unified the API for storage service keyspace tasks (force_keyspace_compaction, force_keyspace_cleanup, upgrade_sstable), making them more consistent and easier to use.
    scylladb#26878

Maintenance

  • Corrected the truncate logic to verify a table’s flush status across all relevant shards, ensuring complete and correct truncation.
    scylladb#27639

  • Ensured the force_keyspace_cleanup_async API correctly handles non-vnode keyspaces, preventing potential misapplication of the cleanup operation.
    scylladb#26715

Raft

  • Ensured the IP-to-ID mapping for a replacing node in raft topology is preserved after restart, maintaining correct cluster membership.
    scylladb#28057

  • Fixed a failed assertion in raft::log::log on single-node cluster restart by improving how the incremental reader selector filters engaged SSTables, enhancing cluster startup reliability.
    scylladb#26247

  • Fixed an issue where a left node with a pending leave request could remain in the raft topology, ensuring the cluster state accurately reflects its members.
    scylladb#27990

  • Revised the topology coordinator to consistently handle raft::request_aborted errors, making the cluster more resilient to Raft-related topology changes.
    scylladb#23589

Reliability

  • Eliminated a race condition between commitlog large allocation and segment termination.
    scylladb#27992

  • Fixed a premature break in tablet_sstable_streamer::stream to ensure all SSTables are processed during streaming, improving the completeness of data transfers.
    scylladb#26979

  • Fixed the tablet scheduler to prevent emitting conflicting migrations for the same tablet across different data centers, ensuring correct data placement and coherence.
    scylladb#26038

  • Prevented a segmentation fault (nullptr dereference) when running concurrent SELECT ... FROM MUTATION_FRAGMENTS(...) and regular SELECT queries on the same partition.
    scylladb#26847

  • Resolved oversized memory allocation errors (seastar_memory) that occurred when creating a large number of tables, allowing for better resource utilization.
    scylladb#26787

  • Synchronized tablet split and load-and-stream processes to eliminate a race condition.
    scylladb#26455

  • Updated tablet scheduler logic to prevent the emission of conflicting migrations (inter-node and intra-node), resolving a potential cause of incorrect reads.
    scylladb#27304

Repair

  • Corrected the repair log format for minimum and maximum tokens to the proper (start, end] notation, improving log clarity and consistency.
    scylladb#27220

  • Hardened batchlog replay during repair to eliminate the severe risk of data resurrection. By ensuring that tablet repair fails if any batch was not successfully sent during the repair process, this fix prevents data previously marked for deletion (via tombstone garbage collection) from reappearing, thus maintaining data integrity and consistency.
    scylladb#24415