[
https://issues.apache.org/jira/browse/GEODE-2683?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15930801#comment-15930801
]
ASF subversion and git services commented on GEODE-2683:
--------------------------------------------------------
Commit 1036259eca5cabd84dbe94da4043b8f266f2b6e6 in geode's branch
refs/heads/develop from zhouxh
[ https://git-wip-us.apache.org/repos/asf?p=geode.git;h=1036259 ]
GEODE-2683: let BR.putAll/removeAll to distribute before notify gateway, which
is the same order as put/destroy
> Lucene query did not match region values
> ----------------------------------------
>
> Key: GEODE-2683
> URL: https://issues.apache.org/jira/browse/GEODE-2683
> Project: Geode
> Issue Type: Bug
> Reporter: xiaojian zhou
> Assignee: xiaojian zhou
> Fix For: 1.2.0
>
>
> There're several root causes. This one is due to the fix in #45782 changed
> the order to notify primary bucket's gateway before distribute to secondary.
> The log is at /export/buglogs_bvt/xzhou/lucene/concParRegHA-0209-235804
> CLIENT vm_1_thr_17_dataStore1_ip-10-32-108-36_11189
> TASK[1] parReg.ParRegTest.HydraTask_HADoEntryOps
> ERROR util.TestException: util.TestException: Lucene query did not match
> region values. missingKeys=[], extraKeys=[Object_13, Object_17, Object_952,
> Object_550, Object_1876, Object_2732, Object_270, Object_4722, Object_4726,
> Object_2537]
> at lucene.LuceneHelper.verifyLuceneIndex(LuceneHelper.java:88)
> at lucene.LuceneTest.verifyLuceneIndex(LuceneTest.java:128)
> at lucene.LuceneTest.verifyFromSnapshotOnly(LuceneTest.java:79)
> at parReg.ParRegTest.verifyFromSnapshot(ParRegTest.java:5638)
> at parReg.ParRegTest.concVerify(ParRegTest.java:6035)
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> at
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
> at
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
> at java.lang.reflect.Method.invoke(Method.java:497)
> at util.MethodCoordinator.executeOnce(MethodCoordinator.java:68)
> at parReg.ParRegTest.HADoEntryOps(ParRegTest.java:2273)
> at parReg.ParRegTest.HydraTask_HADoEntryOps(ParRegTest.java:1032)
> at sun.reflect.GeneratedMethodAccessor24.invoke(Unknown Source)
> at
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
> at java.lang.reflect.Method.invoke(Method.java:497)
> The root cause is:
> T1: A putAll (or removeAll. operation arrived at primary bucket at memberA
> T2: BR.virtualPut() called handleWANEvent() and create shadow key
> T3: PutAll will invoke callback (i.e. write into AEQ) before distribution.
> (Put/Destroy will not have this problem because they distribute before
> callback)
> T4: handleSuccessfulBatchDispatch will send ParallelQueueRemovalMessage to
> the secondary bucket at memberB
> T5: memberB has dataRegion's secondary bucket, but brq is not created yet
> (due to rebalance). So in ParallelQueueRemovalMessage.process(), it will only
> try to remove the event from tempQueue (which does not contain the event, so
> it will do nothing)
> T6: Now, finally the BR.virtualPut()'s distribution arrived at user region's
> secondary bucket at memberB. It will be added into the AEQ (or tempQueue,
> depends).
> T7: memberB becomes new primary (due to rebalance) and re-dispatch the shadow
> key (which has been processed much earlier in memberA). Data mismatch is
> because the replayed event overrides a newer event.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
