[RELEASE] ScyllaDB 6.0.4

Release Notes
, ,
1

The ScyllaDB team announces ScyllaDB Open Source 6.0.4, a bugfix patch release of the ScyllaDB 6.0 stable branch. ScyllaDB Open Source 6.0.4, like all past and future 6.x.y releases, is backward compatible and supports rolling upgrades.

Note that the latest open-source stable branch is 6.1, and you are encouraged to upgrade to it.

Related links:

Issue fixed in this release:

See more in the CQL extension docs. #15559

  • CQL: Allow live update of config enum value via CQL.
    Examples: updating replication_strategy_warn_list
    cqlsh> update system.config set value = '["NetworkTopologyStrategy"]' where name = 'replication_strategy_warn_list';
    #19604
  • Stability: hints: send hints with CL=ALL if target is leaving scylla-dtest#4582
  • Stability (Tablets): Crash when reading specific key using SELECT * FROM MUTATION_FRAGMENTS() from a node which does not own the partition #18786
  • Stability: A race condition between tablet repair and tablet split (the latter happens when a table grows). #19378 #19416.
  • Stability (Tablets): Fatal error during cache update during elasticity test write workload #19873. Root cause is a race between split compaction and tablet migration.
  • Stability: Get rid of fragile compaction group intrusive list #18506
  • Stability: nodetool rebuild --force dc may fail with consistent-topology-changes #20242
  • Stability: nodetool repair command failed with exit code3 during drop keyspace #18490
  • Stability (LWT): paxos_state::guard leaks semaphores resulting in memory leak that leads to large allocations and reactor stalls for LWT users #20602 The issue was introduced in 6.0
  • Stability: Reactor stalls in calculate_schema_digest with 5000 tables #18173
  • Stability: RBNO replace (used by default) is unsafe if the replication factor in a data center is too small. For example with a RF=1, a repair-based replace would not work #16826
  • Stability: repair: dangling reference in insert_repair_meta #20057. May cause an issue in a race between repair and table drop.
  • Stability (repair): repair_service may use batchlog_manager that is not fully initialized #20236
  • Stability: Scylla hangs when it doesn’t have permissions to TLS files #10305
  • Stability (MV): Materialized view updates destined to a node that has left the cluster are now dropped. #19439
  • Stability (Tablets): Tablet split monitor is processing the same table multiple times in a row #20339
  • Stability: Topology coordinator and replacing node stop see each other after entered transition state failing node replace with inter-dc encryption enabled #19025
  • Stability: a potential Use-after-free in alternator TTL impl #19988
  • Stability: db/hints: Make commitlog instances use the commitlog IO scheduling group #20094
  • Metering: Summary, used to report histograms P99 and P95, does not handle histograms infinite bucket gracefully #20255
  • Tracing: raft::request_aborted gives no information what exactly was aborted #18902
  • Performance: Hints can be dropped during decommission #20558
  • Performance: A lock over the node tablet replica map was removed, as it was causing topology changes to be delayed. #18821
  • Install: install.sh installing files with incorrect permission on strict umask setting #14383
  • Admin REST API: A new API system/highest_supported_sstable_version, return the sstable format version supported across the cluster #19772
  • Install: scylla-conf depends on /bin/sh in pretrans #20321
  • Docker: Allow configuring default loglevel with args for container images #19660
  • Tools: stress should use the NetworkTopologyStrategy by default, not SimpleStrategy scylla-tools-java#400
  • EC2: Add i4i instance types to AWS AMI cluster Cloud Formation templates scylla-pkg#4322