Hey all, besides tools like Kopf, what types of things to you do to really dig in and understand if your ES cluster is underpowered? Of course getting results back slower than you would like is a telling sign that you might need to change your shard configuration, index configuration, add additional nodes, etc. What types of things to folks that administer an ES cluster do to identify when the cluster is simply underpowered?
Things we have discussed
- Enable slow log
- Look at # of segments in each index (if we have a ton, should probably optimize if no longer writing to the index correct)?
- We have dedicated masters, client nodes, and data nodes. So none of them are being dual purposed.
Thanks all! There is a good chance we are overlooking some obvious things that seasoned ES admins look at.