We have the new Replica Placement API, which can react to various events like nodeUp, nodeDown etc. Would it make sense to auto rebalance leaders after a certain delay after the last nodeDown/Up event occurred?
Jan > 22. sep. 2022 kl. 05:23 skrev David Smiley <dsmi...@apache.org>: > > SolrCloud's choice of a leader isn't "sticky"; it can change and won't > necessarily change back on its own if the original choice is eligible > again. Setting "preferredLeader" on a replica and calling the > REBALANCELEADERS API is how to make it sticky.[1]. Without this, a rolling > restart (common-place in a cloud-native world?) will leave the last node > without any leaders. That can lead to a resource usage imbalance, > particularly with TLOG due to the higher differentiation of activity a > leader does vs a non-leader. Make sense? > > Would it make sense for collection creation to have a setPreferredLeader > option or some-such boolean name? Or just let the caller/client do so if > it pleases, and maybe mention this as advice in the reference guide? Or > just do this by default for TLOG collection creation... although I wonder > if it'd interfere with how some users manage their clusters? > > [1] > https://solr.apache.org/guide/solr/latest/deployment-guide/collection-management.html#rebalanceleaders > > ~ David Smiley > Apache Lucene/Solr Search Developer > http://www.linkedin.com/in/davidwsmiley --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@solr.apache.org For additional commands, e-mail: dev-h...@solr.apache.org
