ElasticSearch Filter Cache Size is too low and a lot field Eviction

We have ES(1.7.3) cluster with 18 nodes. We have the following settings for each node:

ES_HEAP_SIZE=90G
indices.fielddata.cache.size=75%

When we looked at the monitoring metrics(Elastic HQ), We have "Field Size": ~25GB while the "Filter Cache Size":~5GB. Should the "Filter Cache Size" be 90Gx75%=67.5G?

Under this settings, we are experiencing very slow "Search-Fetch" and "Refresh". And "Field Evictions" are sky high, e.g. some of the nodes have >1000 Field Evictions.

Is there any reasons the filter cache size is not shown as expected? and are we doing something wrong with the configurations?

That's way beyond what we recommend as best practises. Why is it so big?

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