Query latency spikes and Open search contexts

Hi All,
Lately we are seeing latency spikes in production for our queries. We also see
a spike in the number of open search contexts ( it shoots to around 2000 from 400 ). We are using
ES 1.3.9 and it recovers on its own after about 5 minutes. There are no scan/scrolls being used,
so its surprising to see a spike in open search contexts...any clues on why this may be happening?


Are you monitoring things, eg with Marvel or something?
Have you checked your ES logs for GC or anything else?


We are monitoring our ES closely...but so far haven't found much that would help. Any ideas on specific metrics to monitor are most welcome...

Incoming indexing and search requests would be two immediate ones.

Also, upgrade.

Thanks Mark!

We are considering upgrade path to 1.7.3 and then to 2.1.1. just curious, are there any changes with respect to search contexts in versions beyond 1.4.X?


There's a few when you hit 2.X, make sure you run the migration plugin and read breaking changes :slight_smile: