We are suffering lately from very strange problem: nodes API stuck and kibana gives timeout because of it while trying to get nodes list.
Both Kibana and logstash has local ES
6 data nodes with 64 GB of ram and 2TB data
and about 14 indexers
Data pipeline is kafka
There are no logs showing a problem or exception on any ES node. Also once kibana got node list it keeps working well and data is indexed well.
curl -XGET 'http://localhost:9200/_cat/nodes' - stuck
but getting explicit node working well.
2015/08/05 05:07:15 [error] 1694#0: *2397 upstream prematurely closed connection while reading response header from upstream, client: 126.96.36.199, server: kibana.localhost, request: "GET /elasticsearch/nodes?=1438751114353 HTTP/1.1", upstream: "http://127.0.0.1:5601/elasticsearch/nodes?=1438751114353", host: "kibana.eu.ec2ssa.info", referrer: "http://kibana.eu.ec2ssa.info/"