[ https://issues.apache.org/jira/browse/HELIX-50?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13582704#comment-13582704
]
dafu commented on HELIX-50:
---------------------------
restate the problem here.
ideal-state:
replica: 2 (which should be 3)
preference-list: "node_0, node_1, node_2"
current-state:
node_1: slave
node_2: slave
so the best-possible-state will be:
node_0: master
node_1: slave
but since replica is set to 2, we can't bring any instance (node_0 here) to slave state. and
thus helix keep two slave state (but no master state) for the partition.
we need 2 fixes here:
1) validate ideal-state
2) when generate/select message, we take into consideration the state priority, so we should
send offline->slave messages to instances (e.g. node_0) whose desire state (e.g. master)
has a higher priority.
> Ensure num replicas and preference list size in idealstate matches
> ------------------------------------------------------------------
>
> Key: HELIX-50
> URL: https://issues.apache.org/jira/browse/HELIX-50
> Project: Apache Helix
> Issue Type: Bug
> Reporter: kishore gopalakrishna
>
> Currently we dont validate that the number of replicas and size of preference list is
the same. This results in inconsistent behaviour.
--
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
|