[
https://issues.apache.org/jira/browse/ZOOKEEPER-1324?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13647973#comment-13647973
]
Alexander Shraer commented on ZOOKEEPER-1324:
---------------------------------------------
One case where this "if" actually changed behavior is when sid is switching
from being Observer to being Participant. The "if" would see that sid is an
Ovserver in the current config and prevent sid from counting towards the
new-config too, where sid is actually a Participant.
I'm not 100% sure what the right thing to do here. On one hand, acking a
NEWLEADER message means that state transfer is complete, Observer or not, so
why not count this guy's vote. On the other hand, he is not yet a participant,
and in the normal (non-recovery) case we wouldn't count its vote (he wouldn't
send any).
My intuition is that both options are safe in this case since the ACK means
state transfer complete, and for liveness its better to allow it to vote, i.e.,
to remove the if as Flavio now did.
> Remove Duplicate NEWLEADER packets from the Leader to the Follower.
> -------------------------------------------------------------------
>
> Key: ZOOKEEPER-1324
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1324
> Project: ZooKeeper
> Issue Type: Improvement
> Components: quorum
> Affects Versions: 3.5.0
> Reporter: Mahadev konar
> Assignee: Flavio Junqueira
> Priority: Critical
> Fix For: 3.5.0, 3.4.6
>
> Attachments: ZOOKEEPER-1324-branch-3.4.patch, ZOOKEEPER-1324.patch,
> ZOOKEEPER-1324.patch, ZOOKEEPER-1324.patch, ZOOKEEPER-1324.patch,
> ZOOKEEPER-1324.patch, ZOOKEEPER-1324.patch, ZOOKEEPER-1324.patch
>
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
