Task record not created when wait_for_completion=false and server under load

Hi all. I'm experiencing a problem where I'm issuing update_by_query/delete_by_query requests, specifying wait_for_completion=false....and I get back a task id as I should, but when I attempt to GET that task I get resource_not_found_exception. It only happens very sporadically and seemingly only when the cluster is under a good amount of load. Around the same time I am seeing other requests get rejected because the write queue is full (es_rejected_execution_exception) ....is it possible that the update-by-query request has been accepted and is executing, but internally the insertion of the task record was rejected because the write queue was full? It seems like a bug that it would return me a task id that doesn't exist. Does anyone have any theories what's going on here or possible workarounds? Thanks :slight_smile:

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.