[
https://issues.apache.org/jira/browse/SOLR-11011?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16091426#comment-16091426
]
Cao Manh Dat commented on SOLR-11011:
-------------------------------------
Thanks [~shalinmangar]
1. Sure, if we want to use this implementation in the future we can easily move
it to ZkController
2. I think the implementation for {{intToBytes}} and {{bytesToInt}} is trivial,
even we store the number as a string, we still have to convert the string to
bytes, I kinda like current approach.
3. Sure, I will do that.
4. If we throw {{InterruptedException}}, so we have to throw
{{InterruptedException}} in Assign methods too which are propagated to several
other methods.
5. The changes are necessary because whenever we call Assign methods, we will
increase the counter ( then the newReplicaName does not match with precomputed
one ). So I solved the problem by grab replica name from the response.
> Assign.buildCoreName can lead to error in creating a new core when
> legacyCloud=false
> ------------------------------------------------------------------------------------
>
> Key: SOLR-11011
> URL: https://issues.apache.org/jira/browse/SOLR-11011
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Cao Manh Dat
> Assignee: Cao Manh Dat
> Attachments: SOLR-11011.patch, SOLR-11011.patch
>
>
> Here are the case
> {code}
> shard1 : {
> node1 : shard1_replica1,
> node2 : shard1_replica2
> }
> {code}
> node2 go down, autoAddReplicasPlanAction is executed
> {code}
> shard1 : {
> node1 : shard1_replica1,
> node3 : shard1_replica3
> }
> {code}
> node2 back alive, because shard1_replica2 is removed from {{states.json}} so
> that core won't be loaded ( but it won't be removed neither ). Then node1 go
> down, Assign.buildCoreName will create a core with name=shard1_replica2 which
> lead to a failure.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]