The ScyllaDB team announces the availability of ScyllaDB Open Source 2.3.1, a bugfix release of the ScyllaDB Open Source 2.3 stable branch. Release 2.3.1, like all past and future 2.3.y releases, is backward compatible and supports rolling upgrades.
- Get ScyllaDB (Docker, binary packages, and EC2 AMI)
- Get started with ScyllaDB
- Upgrade from 2.x.y to 2.x.z
This maintenance release fixes a critical issue with nodetool cleanup:
nodetool cleanup is used after adding a node to a cluster, to clean partition ranges not owned by a node. The critical issue we found is that nodetool cleanup would wrongly erase up to 2 token ranges that were local to the node. This problem will persist if cleanup is executed on all replicas of this range. The root cause of the issue is #3872, an error in an internal function used by cleanup to identify which token ranges are owned by each node.
If you ran nodetool cleanup, and you have questions about this issue, please let us know via GitHub.
Additional issues solved in ScyllaDB Open Source 2.3.1:
- Gossip: non zero shards may have stale (old) values of gossiper application states for some time #3798. This can create an issue with schema change propagation, for example,
- CQL: some CQL syntax errors can cause ScyllaDB to exit #3740, #3764
- In Transit Encryption: Possible out of memory when using TLS (encryption) with many connections #3757
- CQL: MIN/MAX CQL aggregates were broken for timestamp/timeuuid values. For example
SELECT MIN(date) FROM ks.hashes_by_ruid; where the date is of type timestamp #3789
- CQL: TRUNCATE request could have returned a succeeds response even if it failed on some replicas #3796
- Prometheus: Fix histogram text representation #3827