[ https://issues.apache.org/jira/browse/SOLR-12757?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17399058#comment-17399058 ]
Cassandra Targett commented on SOLR-12757: ------------------------------------------ I think this can be changed since it relates to autoscaling, but I'm not sure if there's some aspect of this that is valid without the autoscaling framework? > Don't load cores that belong to inactive shards > ----------------------------------------------- > > Key: SOLR-12757 > URL: https://issues.apache.org/jira/browse/SOLR-12757 > Project: Solr > Issue Type: Improvement > Components: AutoScaling > Reporter: Andrzej Bialecki > Assignee: Andrzej Bialecki > Priority: Major > > After shard splitting completes and the new sub-shards become active the > original (parent) shards is switched to inactive state. Since 7.4 Solr runs a > scheduled trigger that uses {{InactiveShardPlanAction}} to eventually remove > inactive shards when their TTL expires (2 days by default). > However, in the meantime even though the shard is inactive all cores that > belong to this shard are still running, and should a node restart these cores > would be loaded again too. > Large scale simulated tests show that in case of a constantly growing index > after a while most of the cluster resources are consumed by cores that belong > to inactive shards. > We should change this so that when the switch-over to the new sub-shards is > complete the cores belonging to the inactive parent shard should be unloaded > and not loaded again - at the very least it should be made into an option > that is true by default. -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org For additional commands, e-mail: issues-h...@solr.apache.org