So I'm new to our dev team and I'm trying to trouble shoot our ElasticSearch Performance issues.
I noticed that our data retrieval from the ES instance takes 800ms -1200ms for a result-set as small as 1000 records. I've never seen ES behaving so slowly on a search. Indexing performance is even worse.
I looked at our cluster (which has 4 m4.large.elasticsearch nodes) and realized that we have 70+ indices in our production instance (majority of which are out of date and for back up purposes only). But we have 300+ active primary shards and 700+ active shards when our biggest index doesn't have data bigger than 2 gigs. Each of these indices are set up with a default of 5 shards and 1 replica.
I'm trying to understand (and need some expert opinion so I can go back to my team with findings) if the total number of shards (700+) over the number of nodes (4) will result in the horrible performance we are seeing, and just how bad is the health our cluster (even though the monitor says green).
Thanks,
Daniel