[
https://issues.apache.org/jira/browse/GEODE-5505?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16589099#comment-16589099
]
ASF subversion and git services commented on GEODE-5505:
--------------------------------------------------------
Commit 8ae5808dd7a48eab3be9ec02e61fab0e75002320 in geode's branch
refs/heads/develop from [~bschuchardt]
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=8ae5808 ]
GEODE-5505 Cache listener not invoked on a retried destroy() operation
Delay region synchronization operations to allow clients to retry their
operations. This keeps a server from obtaining the change made by a
client in another server that was only partially distributed. Without
the delay it is possible for the sync to complete before a client retries
its operation, so that the client's replay encounters the state change
it already made in its previous attempt.
With this change a client's replay will either be on a server that didn't
see the previous attempt, or it will be on a server that did see the
previous attempt & has the operation recorded in its EventTracker. In
the first case the operation will be performed again and transmitted
to other servers. In the second case the server will not apply the operation
to its cache but will forward it to other servers.
The delay is based on the server's "maximum time between pings interval",
which defaults to 1 minute.
This closes #2340
> Cache listener not invoked on a retried destroy() operation
> -----------------------------------------------------------
>
> Key: GEODE-5505
> URL: https://issues.apache.org/jira/browse/GEODE-5505
> Project: Geode
> Issue Type: Bug
> Components: regions
> Reporter: Bruce Schuchardt
> Assignee: Bruce Schuchardt
> Priority: Major
> Labels: pull-request-available
> Time Spent: 3h 40m
> Remaining Estimate: 0h
>
> If a server fails while a client is performing a destroy() operation it's
> possible for other servers to have recovered consistency using a distributed
> synchronization operation before the client is able to retry the operation.
> If this happens we accept the destroy() operation in the other servers but
> we're not invoking cache listeners.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
