[
https://issues.apache.org/jira/browse/GEODE-8943?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17291974#comment-17291974
]
Eric Shu edited comment on GEODE-8943 at 3/3/21, 6:12 PM:
----------------------------------------------------------
Compute interested client on remote node (far side from tx host) exists earlier
before 1.14 as register interest can occur after tx committed on tx host.
GEODE-8559 to address a race condition that
LocalRegion.generateLocalFilterRouting can not compute correct interested
clients during tx commit on the remote node.
I do not see a valid way to address this issue without re-compute it again to
avoid data inconsistency between client and server without this change.
was (Author: eshu):
GEODE-8559 to address a race condition that
LocalRegion.generateLocalFilterRouting can not compute correct interested
clients during tx commit.
I do not see a valid way to address this issue without re-compute it again to
avoid data inconsistency between client and server without this change.
> Filterrs (Interest and CQ) is processed multiple times with transactional
> operation on PR
> -----------------------------------------------------------------------------------------
>
> Key: GEODE-8943
> URL: https://issues.apache.org/jira/browse/GEODE-8943
> Project: Geode
> Issue Type: Bug
> Components: cq, regions
> Affects Versions: 1.14.0
> Reporter: Anilkumar Gingade
> Assignee: Eric Shu
> Priority: Major
> Labels: blocks-1.14.0, pull-request-available
>
> The Filters (interest and CQ) could be getting processed multiple times when
> transactional operation is performed.
> By design on PR the filters are processed on primary bucket where the data is
> applied/changed and adjunct message with interested clients are sent to
> remote servers (where the subscription queues are hosted) and replicas; thus
> performing filter processing once and sending the message to remote servers
> only if the filters are satisfied; it looks like currently the filter
> processing is happening both at the primary and secondary buckets for TX
> operation.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
