mirror of
https://github.com/honeymoose/OpenSearch.git
synced 2025-03-09 14:34:43 +00:00
96f4a72825
Submit job updates to a concurrent queue when job update has been processed by ClusterService. Then from a background thread delegate the job updates to the node running the autodetect process. This maintains the same order as how the job config updates have occurred to the cluster state and thus preventing job config updates to the same job to arrive in the wrong order to the job's autodetect process. (the expectation is that in practise this will rarely happen) The behaviour of the update api changes with this pr, because the api now returns when the update has been made to cluster state, whereas before it would return when the update was made to the autodetect process too. Updating the autodetect process happens in the background. I think that this change in behaviour is acceptable. Use TP#scheduleWithFixedDelay(...) instead of TP#schedule(...) and removed the custom rescheduling and cancelling. Also changed LocalNodeMasterListener#executorName to SAME Original commit: elastic/x-pack-elasticsearch@c24c0dd7d7