Elasticsearch 1.7.4 only one node queue is full other is normal, maybe is a bug

[whybangbang]

our's es' version is 1.7.4 ,today es-6 is down and is unavailable ,es-6'shard move to es-7 ,at this moment ,

[es-m-7] [14297421] Failed to execute fetch phase
org.elasticsearch.common.util.concurrent.EsRejectedExecutionException: rejected execution (queue capacity 1000)

es-7' load is high but other node is normal and search queue is low

i think maybe is a bug,

[warkolm (Mark Walkom)]

Why do you think it is a bug? Can you give us steps to replicate this behaviour to check if it is a bug?

Are you monitoring your hosts to see what is happening and why one is over loaded?

[whybangbang]

es-7 and es-5 is in a common cluster and have same number sahrd, but at 03:36 es-7'queue is full ,but es-5 is normal

in normal case, search request should allocate to every node ，but in this monitor graph，all request route to es-7

@warkolm

[warkolm (Mark Walkom)]

What's sending the searches to your cluster then?