I'm attempting to diagnose an increase in overall latency and error rates
during periods of high traffic. I'm looking at my thread_pool settings
right now.
Here are my node settings (_cluster/nodes?clear=true&thread_pool=true):
Specifically, please note thread_pool.search.type: fixed, and
thread_pool.search.capacity: -1. I understand this means an unbounded
queue, which means no request may be rejected, merely delayed.
Here are my node stats (_cluster/nodes/stats?clear=true&thread_pool=true):
Specifically, please look at thread_pool.search.rejected:
And the same output a few minutes later:
What is causing these search requests to get rejected? From my
understanding, this should be impossible. Or, is this a red herring?
I'm looking at this gist and it looks like you issued this command on an older version of ES. I am on 1.7.3 and this API looks gone. Has it been replaced with another command?
I am trying to clear my search queue as it is jammed up with 6000+ queries.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.