Top 10 Elasticsearch Mistakes

Last updated on Dec 10, 2017

1. Upgrading to the new major version right after its release without waiting for the inevitable .1 release
2. Remembering that you said, “We don’t need backups, we have shard replicas” to your manager during an 8-hour cluster recovery session
3. Not running dedicated masters and wondering why your whole cluster becomes unresponsive during high load
4. In a room full of Elasticsearch fans suggest that Elasticsearch should use ZooKeeper like SolrCloud and avoid split-brain
5. Running a single master and wondering why it takes the whole cluster down with it
6. Running a significant terms aggregation on an analyzed field and wondering where all the memory/heap went
7. Not using G1 GC with large heaps because Robert Muir claims G1 and Lucene/Elasticsearch don’t get along (just kidding, Robert!)
8. Giving Elasticsearch JVM 32 GB heap and thinking you’re so clever ‘cause you’re still using CompressedOops. Tip: you ain’t
9. Restarting multiple nodes too fast without waiting for the cluster to go green between node restarts
10. …and last but not least: not taking Sematext Elasticsearch guru @radu0gheorghe’s upcoming Elasticsearch / ELK Stack Online Training courses, covering: Intro to Elasticsearch, Elasticsearch Operations and Elasticsearch for Logging.