Today we are pleased to announce the bug fix release of Elasticsearch 5.0.2, based on Lucene 6.2.1. This is the latest stable release and is already available for deployment on Elastic Cloud, our Elasticsearch-as-a-service platform. All users are advised to upgrade.
Latest stable release:
Full details of the changes in this release are available in the release notes listed above, but there are a few important changes which are worth singling out:
- A bug fix in X-Pack v5.0.1 introduced a new bug which prevented users of Security from doing a rolling upgrade between 5.0.0 and 5.0.1. Users of X-Pack 5.0.0 and 5.0.1 can now safely do a rolling upgrade to 5.0.2.
- A bug in caching resulted in frequent failures under load when “running-as” a native user in X-Pack Security.
- Cluster instability and master-loss combined with long-running snapshots could result in primary shards becoming unallocated and unallocatable. This is fixed by cancelling any ongoing snapshots on an isolated node which rejoins the cluster, plus changing a shard-locked exception from a hard failure (never retried) to a soft failure.
- Cluster state update tasks which timed out could lead to memory leaks on the master, eventually resulting in OOM.
- Elasticsearch should “die with dignity” when an OOM exception is thrown, but these exceptions were being swallowed by the network layer and by Lucene.
- The Transport Client now has a HostFailureListener which is called when the client fails to connect to a node, allowing the application to update the list of nodes known by the client.
- The Tribe node can now be configured to use custom ports instead of the defaults.