[
https://issues.apache.org/jira/browse/ZOOKEEPER-1699?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexander Shraer updated ZOOKEEPER-1699:
----------------------------------------
Attachment: ZOOKEEPER-1699-v4.patch
Hi Flavio,
I actually like a lot that we can reuse the Proposal logic here and not
duplicate it. Its not only the hasallquorums, but also the fact that we have
potentially multiple quorumverifiers and we need to have a separate ackSet for
each one, such that only appropriate voters have an ack in the appropriate
ackSet. The reason is that the normal quorumverifier check for majority only
checks that the number of votes is more than half, it doesn't check the votes
themselves - we have to do that before adding the acks.
I changed the name as you suggest.
Thanks,
Alex
> Leader should timeout and give up leadership when losing quorum of last
> proposed configuration
> ----------------------------------------------------------------------------------------------
>
> Key: ZOOKEEPER-1699
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1699
> Project: ZooKeeper
> Issue Type: Bug
> Components: server
> Affects Versions: 3.5.0
> Reporter: Alexander Shraer
> Assignee: Alexander Shraer
> Priority: Blocker
> Fix For: 3.5.0
>
> Attachments: ZOOKEEPER-1699-v1.patch, ZOOKEEPER-1699-v2.patch,
> ZOOKEEPER-1699-v3.patch, ZOOKEEPER-1699-v4.patch, ZOOKEEPER-1699.patch
>
>
> A leader gives up leadership when losing a quorum of the current
> configuration.
> This doesn't take into account any proposed configuration. So, if
> a reconfig operation is in progress and a quorum of the new configuration is
> not
> responsive, the leader will just get stuck waiting for it to ACK the reconfig
> operation, and will never timeout.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
