You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@curator.apache.org by "Anders Wallgren (JIRA)" <ji...@apache.org> on 2013/06/27 01:52:19 UTC
[jira] [Updated] (CURATOR-44) LeaderSelector does not assign leader
randomly
[ https://issues.apache.org/jira/browse/CURATOR-44?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Anders Wallgren updated CURATOR-44:
-----------------------------------
Description:
The javadoc for LeaderSelector says leadership is assigned randomly but this does not appear to be the case.
A fair mutex is used so it appears that leaders are selected in the order that they attempt to acquire the mutex.
This probably wouldn't be much more than a documentation issue except that it tends to concentrate leaders on one node in the cluster when each node participates in more than one election (and assuming the node starts the LeaderSelectors right around the time the node comes up, which is the case in our usage).
was:
The javadoc for LeaderSelector says leadership is assigned randomly but this does not appear to be the case.
A fair mutex is used so it appears that leaders are selected in the order that they attempt to acquire the mutex.
This probably wouldn't be much more than a documentation issue except that it tends to concentrate leaders on one node in the cluster when each node participates in more than one election and tends to start the LeaderSelectors right around the time the node comes up.
> LeaderSelector does not assign leader randomly
> ----------------------------------------------
>
> Key: CURATOR-44
> URL: https://issues.apache.org/jira/browse/CURATOR-44
> Project: Apache Curator
> Issue Type: Bug
> Components: Recipes
> Affects Versions: 2.0.1-incubating
> Reporter: Anders Wallgren
>
> The javadoc for LeaderSelector says leadership is assigned randomly but this does not appear to be the case.
> A fair mutex is used so it appears that leaders are selected in the order that they attempt to acquire the mutex.
> This probably wouldn't be much more than a documentation issue except that it tends to concentrate leaders on one node in the cluster when each node participates in more than one election (and assuming the node starts the LeaderSelectors right around the time the node comes up, which is the case in our usage).
--
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