[
https://issues.apache.org/jira/browse/GEODE-7702?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17103961#comment-17103961
]
Xiaojian Zhou commented on GEODE-7702:
--------------------------------------
It looks like my fix to removeAll part changed the current behavior and caused
the DistributedNoAckRegionCCEDUnitTest.
testConcurrentEventsOnNonReplicatedRegion occasionally fail.
The root cause is: the test purposely used no-ack and put a put then destroy on
a normal member. That means the put's distribution of the put is still on going
when the destroy happened. In old behavior, the destroy will be on local member
then do distribution because local member has the key.
But my fix forced the normal member to always do remote destroy. This caused
the sideeffect on no-ack case.
After I think over, my previous fix to removeAll from normal member is
unnecessary. If the clear and the removeAll from normal mis-ordered, the result
will be the same.
So I will revert my previous fix and refine it.
> DistributedAckRegionCCEDUnitTest >
> testClearOnNonReplicateWithConcurrentEvents is showing a product bug
> -------------------------------------------------------------------------------------------------------
>
> Key: GEODE-7702
> URL: https://issues.apache.org/jira/browse/GEODE-7702
> Project: Geode
> Issue Type: Bug
> Components: client/server
> Reporter: Mark Hanson
> Assignee: Xiaojian Zhou
> Priority: Major
> Labels: GeodeCommons
>
> testClearOnNonReplicateWithConcurrentEvents in
> DistributedAckRegionCCEDUnitTest.java
> versionTestClearOnNonReplicateWithConcurrentEvents in
> MultiVMRegionTestCase.java
> doOpsLoop
> doOpsLoopNoFlush
> {noformat}
> case 5:
> if (includeClear) {
> CCRegion.clear();
> break;
> } else {
> if (CCRegion.getAttributes().getDataPolicy().withReplication()) {
> if (oldkey != null) {
> CCRegion.putIfAbsent(oldkey, value);
> }
> break;
> } // else fall through to invalidate
> } {noformat}
> the addition of this chunk of code causes this test to fail.
> The core of the problem is that a putall and a clear are happening
> concurrently and the "system" does not respond by either clearing all entries
> or letting all entries persist.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
