[
https://issues.apache.org/jira/browse/SOLR-9555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16138255#comment-16138255
]
Cao Manh Dat commented on SOLR-9555:
------------------------------------
Great work [~mdrob], but there are one thing that I concern about your
direction.
Firstly, There are one requirement for LIR implementation is : the replica that
miss the update, should not allowed to serve query requests.
- In current implementation, Leader will publish the replica as DOWN, so all
clients will know that and stop query that replica. ( low latency )
- In your implementation, Leader only publish LIR state of that replica as
DOWN, so all clients will keep query that replica, until the replica
acknowledge the change in LIR node then publish it self as DOWN ( high latency
)
Secondly, Now we have 2 nodes update state of a replica which is considered as
bad practice. With your implementation, we have 2 nodes update lir state of a
replica, will we have race condition bug in the future?
> Leader incorrectly publishes state for replica when it puts replica into LIR.
> -----------------------------------------------------------------------------
>
> Key: SOLR-9555
> URL: https://issues.apache.org/jira/browse/SOLR-9555
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Alan Woodward
> Attachments: lir-9555-problem.png, SOLR-9555.patch, SOLR-9555.patch,
> SOLR-9555.patch, SOLR-9555-WIP-2.patch, SOLR-9555-WIP-3.patch,
> SOLR-9555-WIP.patch
>
>
> See
> https://jenkins.thetaphi.de/job/Lucene-Solr-master-Linux/17888/consoleFull
> for an example
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
