[ https://issues.apache.org/jira/browse/SOLR-5310?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13807797#comment-13807797 ]
ASF subversion and git services commented on SOLR-5310: ------------------------------------------------------- Commit 1536608 from [~noble.paul] in branch 'dev/branches/branch_4x' [ https://svn.apache.org/r1536608 ] SOLR-5311 - Avoid registering replicas which are removed , SOLR-5310 -Add a collection admin command to remove a replica > Add a collection admin command to remove a replica > -------------------------------------------------- > > Key: SOLR-5310 > URL: https://issues.apache.org/jira/browse/SOLR-5310 > Project: Solr > Issue Type: Improvement > Components: SolrCloud > Reporter: Noble Paul > Assignee: Noble Paul > Attachments: SOLR-5310.patch, SOLR-5310.patch > > Original Estimate: 72h > Remaining Estimate: 72h > > the only way a replica can removed is by unloading the core .There is no way > to remove a replica that is down . So, the clusterstate will have > unreferenced nodes if a few nodes go down over time > We need a cluster admin command to clean that up > e.g: > /admin/collections?action=DELETEREPLICA&collection=coll1&shard=shard1&replica=core_node3 > The system would first see if the replica is active. If yes , a core UNLOAD > command is fired , which would take care of deleting the replica from the > clusterstate as well > if the state is inactive, then the core or node may be down , in that case > the entry is removed from cluster state -- This message was sent by Atlassian JIRA (v6.1#6144) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org