[
https://issues.apache.org/jira/browse/ZOOKEEPER-1516?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13421336#comment-13421336
]
Flavio Junqueira commented on ZOOKEEPER-1516:
---------------------------------------------
If you reduce that value, you increase the probability of the algorithm not
converging. You'll make your deployment unstable. Given that leader election
shouldn't be a common event, 200ms seems pretty acceptable to me in general and
I don't recommend allowing administrators to calibrate that value.
> Configurable finalizeWait for FastLeaderElection
> ------------------------------------------------
>
> Key: ZOOKEEPER-1516
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1516
> Project: ZooKeeper
> Issue Type: Improvement
> Components: leaderElection, quorum, server
> Affects Versions: 3.3.5
> Environment: Gentoo linux, any environment is affected.
> Reporter: Ian Babrou
> Labels: performance
>
> FastLeaderElection has final static int finalizeWait = 200. This is time to
> wait after successful leader election. I don't know what could happen, but
> 200ms is too slow for production environment under heavy load.
> I changed it to 20ms and everything still works for me.
> I propose to make this value configurable with default value of 200 to not
> affect current installations.
> Combined with #ZOOKEEPER-1515 it could improve leader election and make it
> 10x times faster: 1500ms -> 180ms including 100ms for 2 faileed new leader
> connections.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
