Have a strange issue where we have around 1500 active tasks on our 5 node 5.2 ES cluster. They are all on the same node (current master, and endpoint for Kibana/Monitoring) and go back for days.
Any ideas what these could be? I tried to do a bulk cancel on them to no avail.
curl "http://<server>:9200/_cat/tasks?v&h=action,task_id,parent_task_id,type,timestamp"
... indices:admin/get I_eYf5eiTX-uwkuoO2huEA:26400939 - transport 09:58:17
...
I'm not sure if it's causing an issue, but it's not happening on our 2.3 cluster.
Not quite sure what would constitute as a transport client. We do use the readonlyrest plugin though, although that's on all nodes in the cluster, not just the master (where I'm seeing this).
I've done a bit more digging on this - it appears to be related to http auth, readonlyrest and Sense. When you are prompted for user/password it's logged as a failed attempt in the ROR logs, but the indices:admin/get task hangs around.
I've raised it to readonlyrest but I'm not sure if this is an ES bug or ROR - the problem is I can't cancel the tasks in ES.
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.