I have an Elasticsearch cluster with about 10 slaves nodes. Nodes are c4.xlarge (16 core and 32G ram). When run a term aggregation query (it's actually aggregates on an analyzed field), it takes about 2 minutes to get result back and I see there are a lot of fielddata eviction. Usually the query finishes in less than 10 seconds.
I have following config in my both cluster:
indices.fielddata.cache.size: 50% indices.breaker.fielddata.limit: 70% indices.breaker.request.limit: 30%
Is there any direction or clue I can look for? I don't really have a good idea to debug this. I know fielddata is used for filter but no more than that. Any tip will help. Appreciate it!