[
https://issues.apache.org/jira/browse/GEODE-10242?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17528942#comment-17528942
]
ASF subversion and git services commented on GEODE-10242:
---------------------------------------------------------
Commit d6f4e1f61aad5bcaac8d27b100a80ea645f53bf4 in geode's branch
refs/heads/feature/GEODE-10242 from Eric Shu
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=d6f4e1f61a ]
GEODE-10242: Do not release primary lock prematurely
* When depose primary during rebalance, do not release the primary lock
before all colocated child buckets has deposed primary. This is to
ensure that the node becomes new primary can only acquire the primary
lock afterwards.
* All colocated buckets now share the same primaryMoveReadWriteLock.
When parent bucket is being moved, no operations will be executed on
child buckets as well. So moving primary for all colocated buckets
shold be faster, and there is no need to hold parent locks anymore.
> Same tail key can be generated for different events on (different colocated
> regions) from different servers
> -----------------------------------------------------------------------------------------------------------
>
> Key: GEODE-10242
> URL: https://issues.apache.org/jira/browse/GEODE-10242
> Project: Geode
> Issue Type: Bug
> Components: wan
> Affects Versions: 1.12.0
> Reporter: Eric Shu
> Assignee: Eric Shu
> Priority: Major
> Labels: GeodeOperationAPI, blocks-1.15.0, pull-request-available
>
> For colocated partition regions with parallel wan queue, rebalance could
> cause primary to be moved. This can lead to a case that one server has the
> primary bucket for the parent region, but another has the primary bucket for
> the child region. As colocated regions show the same parallel wan queue, both
> server will generate next tail key for different events. This will cause some
> event not dispatched to the other wan site and thus event lost.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
