Best non analyzed aggregation setup to avoid 30 second timeout

Hey guys,

I'm using Elasticsearch and Kibana 5 for data analytics primarily.

Brief overview of my setup

  • ES has 1 Node with 12 GB memory
  • 300 million indexed documents across approx 300 indexes with 1 shard each (no replicas)
  • 100GB indexed data

I don't need to analyze any of the string fields at all. They are just meta data used for filtering and aggregations.

Once I start filtering on > a years worth of data I frequently encounter the 30 second timeout on Kibana dashboards.

I was wondering how I could improve my setup to improve response times and avoid this timeout without increasing the timeout value(which wouldn't solve the problem). I suspect my setup is a little under provisioned

I was thinking of adding 2 more nodes and seeing what happens.

Does that sound sensible?

I was also wondering if theres any index config I could do that would improve performance, do I need to switch to doc_values? or are they enabled by default? I suspect they are.

Any suggestions would be greatly appreciated.

Regards,
am87

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.