[ https://issues.apache.org/jira/browse/SOLR-12088?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16399608#comment-16399608 ]
Jerry Bao commented on SOLR-12088: ---------------------------------- We've been running on Solr 7.2.1, so its all been state.json and not clusterstate.json. In regards to re-issuing the DELETEREPLICA command, sometimes that fails and I filed a Jira for that here: SOLR-12087. That was what was causing this second issue here. For example purposes, our indexing latency went from 2s to 1.7s after deleting the dead replicas. One thing I did notice is that the dead replicas spam the logs with "unable to unload non-existent core" on the machine that hosts the dead replicas. Could be a side affect? > Shards with dead replicas cause increased write latency > ------------------------------------------------------- > > Key: SOLR-12088 > URL: https://issues.apache.org/jira/browse/SOLR-12088 > Project: Solr > Issue Type: Bug > Security Level: Public(Default Security Level. Issues are Public) > Components: SolrCloud > Affects Versions: 7.2 > Reporter: Jerry Bao > Priority: Major > > If a collection's shard contains dead replicas, write latency to the > collection is increased. For example, if a collection has 10 shards with a > replication factor of 3, and one of those shards contains 3 replicas and 3 > downed replicas, write latency is increased in comparison to a shard that > contains only 3 replicas. > My feeling here is that downed replicas should be completely ignored and not > cause issues to other alive replicas in terms of write latency. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org