[jira] [Commented] (SOLR-12176) Improve FORCELEADER to handle the case when a replica win the election but does not present in clusterstate
[ https://issues.apache.org/jira/browse/SOLR-12176?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16424947#comment-16424947 ] ASF subversion and git services commented on SOLR-12176: Commit 34b83ed86993d71ba3bb9ae58a3df8ce4351a045 in lucene-solr's branch refs/heads/master from [~caomanhdat] [ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=34b83ed ] SOLR-12176: Improve FORCELEADER to handle the case when a replica win the election but does not present in clusterstate > Improve FORCELEADER to handle the case when a replica win the election but > does not present in clusterstate > --- > > Key: SOLR-12176 > URL: https://issues.apache.org/jira/browse/SOLR-12176 > Project: Solr > Issue Type: Improvement > Security Level: Public(Default Security Level. Issues are Public) >Reporter: Cao Manh Dat >Assignee: Cao Manh Dat >Priority: Major > Attachments: SOLR-12176.patch > > > There can be the case when a replica wins the election but it does not > present in clusterstate. Maybe when the Overseer sent the UNLOAD request to > the LEADER (in DeleteReplicaCmd), it met some exception (therefore the > request never reach the LEADER), the Overseer it that case will forcefully > remove the LEADER from clusterstate. > If a shard reaches that case, users will only see a leaderless shard and call > FORCELEADER won't be able to solve their problem. Therefore FORCELEADER > should be more robust, to handle such cases. -- 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
[jira] [Commented] (SOLR-12176) Improve FORCELEADER to handle the case when a replica win the election but does not present in clusterstate
[ https://issues.apache.org/jira/browse/SOLR-12176?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16424948#comment-16424948 ] ASF subversion and git services commented on SOLR-12176: Commit 3c68f3d63769ec1e9c7400a0974837f051046a65 in lucene-solr's branch refs/heads/branch_7x from [~caomanhdat] [ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=3c68f3d ] SOLR-12176: Improve FORCELEADER to handle the case when a replica win the election but does not present in clusterstate > Improve FORCELEADER to handle the case when a replica win the election but > does not present in clusterstate > --- > > Key: SOLR-12176 > URL: https://issues.apache.org/jira/browse/SOLR-12176 > Project: Solr > Issue Type: Improvement > Security Level: Public(Default Security Level. Issues are Public) >Reporter: Cao Manh Dat >Assignee: Cao Manh Dat >Priority: Major > Attachments: SOLR-12176.patch > > > There can be the case when a replica wins the election but it does not > present in clusterstate. Maybe when the Overseer sent the UNLOAD request to > the LEADER (in DeleteReplicaCmd), it met some exception (therefore the > request never reach the LEADER), the Overseer it that case will forcefully > remove the LEADER from clusterstate. > If a shard reaches that case, users will only see a leaderless shard and call > FORCELEADER won't be able to solve their problem. Therefore FORCELEADER > should be more robust, to handle such cases. -- 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
[jira] [Commented] (SOLR-12176) Improve FORCELEADER to handle the case when a replica win the election but does not present in clusterstate
[ https://issues.apache.org/jira/browse/SOLR-12176?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16424913#comment-16424913 ] Cao Manh Dat commented on SOLR-12176: - I will commit soon if no one has any objection. > Improve FORCELEADER to handle the case when a replica win the election but > does not present in clusterstate > --- > > Key: SOLR-12176 > URL: https://issues.apache.org/jira/browse/SOLR-12176 > Project: Solr > Issue Type: Improvement > Security Level: Public(Default Security Level. Issues are Public) >Reporter: Cao Manh Dat >Assignee: Cao Manh Dat >Priority: Major > Attachments: SOLR-12176.patch > > > There can be the case when a replica wins the election but it does not > present in clusterstate. Maybe when the Overseer sent the UNLOAD request to > the LEADER (in DeleteReplicaCmd), it met some exception (therefore the > request never reach the LEADER), the Overseer it that case will forcefully > remove the LEADER from clusterstate. > If a shard reaches that case, users will only see a leaderless shard and call > FORCELEADER won't be able to solve their problem. Therefore FORCELEADER > should be more robust, to handle such cases. -- 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
[jira] [Commented] (SOLR-12176) Improve FORCELEADER to handle the case when a replica win the election but does not present in clusterstate
[ https://issues.apache.org/jira/browse/SOLR-12176?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16423724#comment-16423724 ] Cao Manh Dat commented on SOLR-12176: - Zombie leader is a leader won the election (created "/collections/leaders/shard/leader") but does not exist in clusterstate. The idea of the patch is finding such zombie leader and unload it. > Improve FORCELEADER to handle the case when a replica win the election but > does not present in clusterstate > --- > > Key: SOLR-12176 > URL: https://issues.apache.org/jira/browse/SOLR-12176 > Project: Solr > Issue Type: Improvement > Security Level: Public(Default Security Level. Issues are Public) >Reporter: Cao Manh Dat >Assignee: Cao Manh Dat >Priority: Major > Attachments: SOLR-12176.patch > > > There can be the case when a replica wins the election but it does not > present in clusterstate. Maybe when the Overseer sent the UNLOAD request to > the LEADER (in DeleteReplicaCmd), it met some exception (therefore the > request never reach the LEADER), the Overseer it that case will forcefully > remove the LEADER from clusterstate. > If a shard reaches that case, users will only see a leaderless shard and call > FORCELEADER won't be able to solve their problem. Therefore FORCELEADER > should be more robust, to handle such cases. -- 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
