[jira] [Commented] (SOLR-10879) DELETEREPLICA and DELETENODE commands should prevent data loss when replicationFactor==1
[ https://issues.apache.org/jira/browse/SOLR-10879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16074073#comment-16074073 ] ASF subversion and git services commented on SOLR-10879: Commit 0324da8289e148d627b9a45c3105bab6ed7573e6 in lucene-solr's branch refs/heads/branch_7_0 from [~ab] [ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=0324da8 ] SOLR-10879: Make sure we don't lose single replicas when deleting a node. > DELETEREPLICA and DELETENODE commands should prevent data loss when > replicationFactor==1 > > > Key: SOLR-10879 > URL: https://issues.apache.org/jira/browse/SOLR-10879 > Project: Solr > Issue Type: Bug > Security Level: Public(Default Security Level. Issues are Public) >Affects Versions: 6.6, 7.0, 6.7 >Reporter: Andrzej Bialecki >Assignee: Andrzej Bialecki > > There should be some level of protection against inadvertent data loss when > issuing these commands when replicationFactor is 1 - deleting a node or a > replica in this case will be equivalent to completely deleting some shards. > This is further complicated by the replica types - there could be still > remaining replicas after the operation, but if they are all of PULL type then > none of them will ever become a shard leader. > We could require that the command should fail in such case unless a boolean > option "force==true" is specified. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-10879) DELETEREPLICA and DELETENODE commands should prevent data loss when replicationFactor==1
[ https://issues.apache.org/jira/browse/SOLR-10879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16074072#comment-16074072 ] ASF subversion and git services commented on SOLR-10879: Commit 30352e72505dd33901158bf8fc76aa98861ab8cc in lucene-solr's branch refs/heads/branch_7x from [~ab] [ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=30352e7 ] SOLR-10879: Make sure we don't lose single replicas when deleting a node. > DELETEREPLICA and DELETENODE commands should prevent data loss when > replicationFactor==1 > > > Key: SOLR-10879 > URL: https://issues.apache.org/jira/browse/SOLR-10879 > Project: Solr > Issue Type: Bug > Security Level: Public(Default Security Level. Issues are Public) >Affects Versions: 6.6, 7.0, 6.7 >Reporter: Andrzej Bialecki >Assignee: Andrzej Bialecki > > There should be some level of protection against inadvertent data loss when > issuing these commands when replicationFactor is 1 - deleting a node or a > replica in this case will be equivalent to completely deleting some shards. > This is further complicated by the replica types - there could be still > remaining replicas after the operation, but if they are all of PULL type then > none of them will ever become a shard leader. > We could require that the command should fail in such case unless a boolean > option "force==true" is specified. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-10879) DELETEREPLICA and DELETENODE commands should prevent data loss when replicationFactor==1
[ https://issues.apache.org/jira/browse/SOLR-10879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16074070#comment-16074070 ] ASF subversion and git services commented on SOLR-10879: Commit cb23fa9b4efa5fc7c17f215f507901d459e9aa6f in lucene-solr's branch refs/heads/master from [~ab] [ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=cb23fa9 ] SOLR-10879: Make sure we don't lose single replicas when deleting a node. > DELETEREPLICA and DELETENODE commands should prevent data loss when > replicationFactor==1 > > > Key: SOLR-10879 > URL: https://issues.apache.org/jira/browse/SOLR-10879 > Project: Solr > Issue Type: Bug > Security Level: Public(Default Security Level. Issues are Public) >Affects Versions: 6.6, 7.0, 6.7 >Reporter: Andrzej Bialecki >Assignee: Andrzej Bialecki > > There should be some level of protection against inadvertent data loss when > issuing these commands when replicationFactor is 1 - deleting a node or a > replica in this case will be equivalent to completely deleting some shards. > This is further complicated by the replica types - there could be still > remaining replicas after the operation, but if they are all of PULL type then > none of them will ever become a shard leader. > We could require that the command should fail in such case unless a boolean > option "force==true" is specified. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-10879) DELETEREPLICA and DELETENODE commands should prevent data loss when replicationFactor==1
[
https://issues.apache.org/jira/browse/SOLR-10879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16048136#comment-16048136
]
Tomás Fernández Löbbe commented on SOLR-10879:
--
OK, I see that this validation is only done when you delete replicas by count,
not if you delete replicas by name (which is what {{DeleteNodeCmd}} does)
> DELETEREPLICA and DELETENODE commands should prevent data loss when
> replicationFactor==1
>
>
> Key: SOLR-10879
> URL: https://issues.apache.org/jira/browse/SOLR-10879
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
>Affects Versions: 6.6, master (7.0), 6.7
>Reporter: Andrzej Bialecki
>Assignee: Andrzej Bialecki
>
> There should be some level of protection against inadvertent data loss when
> issuing these commands when replicationFactor is 1 - deleting a node or a
> replica in this case will be equivalent to completely deleting some shards.
> This is further complicated by the replica types - there could be still
> remaining replicas after the operation, but if they are all of PULL type then
> none of them will ever become a shard leader.
> We could require that the command should fail in such case unless a boolean
> option "force==true" is specified.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-10879) DELETEREPLICA and DELETENODE commands should prevent data loss when replicationFactor==1
[
https://issues.apache.org/jira/browse/SOLR-10879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16048118#comment-16048118
]
Tomás Fernández Löbbe commented on SOLR-10879:
--
Isn't this what {{DeleteReplicaCmd#validateReplicaAvailability(...)}} is doing?
{code:java}
if (allReplicasForShard.size() == 1) {
throw new SolrException(SolrException.ErrorCode.BAD_REQUEST, "There is
only one replica available in shard/collection: " +
shard + "/" + collectionName + ". Cannot delete that.");
}
{code}
> DELETEREPLICA and DELETENODE commands should prevent data loss when
> replicationFactor==1
>
>
> Key: SOLR-10879
> URL: https://issues.apache.org/jira/browse/SOLR-10879
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
>Affects Versions: 6.6, master (7.0), 6.7
>Reporter: Andrzej Bialecki
>Assignee: Andrzej Bialecki
>
> There should be some level of protection against inadvertent data loss when
> issuing these commands when replicationFactor is 1 - deleting a node or a
> replica in this case will be equivalent to completely deleting some shards.
> This is further complicated by the replica types - there could be still
> remaining replicas after the operation, but if they are all of PULL type then
> none of them will ever become a shard leader.
> We could require that the command should fail in such case unless a boolean
> option "force==true" is specified.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
