You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@zookeeper.apache.org by "Benton Liang (Jira)" <ji...@apache.org> on 2022/09/30 22:24:00 UTC
[jira] [Updated] (ZOOKEEPER-4620) zookeeper leader election time metric not reported correctly
[ https://issues.apache.org/jira/browse/ZOOKEEPER-4620?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Benton Liang updated ZOOKEEPER-4620:
------------------------------------
Description:
We added support to pro-actively give up leadership when a node is going down. With this the leader nominates the next leader before giving up leader ship. This avoids the leadership election. There is a bug where they calculate the election time. Normally when the leadership election starts, the start time is noted, and after the election the delta between the current time and startup time is used to determine the leadership time. But now since we are avoiding the leadership election, the start time is not updated this shows up as a big delta.
example:
{code:java}
FOLLOWING - LEADER ELECTION TOOK - 1206674579 MS{code}
was:
We added support to pro-actively give up leadership when a node is going down. With this the leader nominates the next leader before giving up leader ship. This avoids the leadership election. There is a bug where they calculate the election time. Normally when the leadership election starts, the start time is noted, and after the election the delta between the current time and startup time is used to determine the leadership time. But now since we are avoiding the leadership election, the start time is not updated this shows up as a big delta.
example: `FOLLOWING - LEADER ELECTION TOOK - 1206674579 MS`
> zookeeper leader election time metric not reported correctly
> ------------------------------------------------------------
>
> Key: ZOOKEEPER-4620
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-4620
> Project: ZooKeeper
> Issue Type: Bug
> Components: leaderElection, metric system
> Reporter: Benton Liang
> Priority: Minor
>
> We added support to pro-actively give up leadership when a node is going down. With this the leader nominates the next leader before giving up leader ship. This avoids the leadership election. There is a bug where they calculate the election time. Normally when the leadership election starts, the start time is noted, and after the election the delta between the current time and startup time is used to determine the leadership time. But now since we are avoiding the leadership election, the start time is not updated this shows up as a big delta.
> example:
> {code:java}
> FOLLOWING - LEADER ELECTION TOOK - 1206674579 MS{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)