Elasticsearch highly unstable since upgrade to 5.2.0
Elasticsearch has been crashing regularly since the upgrade to 5.2.0. No changes have been made to the settings since prior to the upgrade. My initial guess is that this is a Lucene bug: https://issues.apache.org/jira/browse/LUCENE-7647
The server remains stable for roughly 24 hours before killing ES with an Out of Memory error. After that, restarting ES will only result in the same error over and over until the server it rebooted. Then we get another 24 hours.
I had made some changes on the log server to improve Logstash performance but those were reverted yesterday and we're still seeing the same issues.