jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benjamin Papez (JIRA)" <j...@apache.org>
Subject [jira] Commented: (JCR-2357) Duplicate entries in the index
Date Tue, 20 Oct 2009 08:43:59 GMT

    [ https://issues.apache.org/jira/browse/JCR-2357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12767750#action_12767750
] 

Benjamin Papez commented on JCR-2357:
-------------------------------------

So a solution for this problem would either be that the SearchManager always is the first
event consumer (maybe no good, because too stringent), OR that it is considering nested SearchManager.onEvent
executions triggered via session.save in preceding event consumers, which can already create
index entries for nodes, where the NODE_ADDED event is not yet handled by the SearchManager.onEvent
higher up the stack. For cases of NODE_ADDED events you do not try (probably due to performance
reasons ?) to remove already existing entries in the index, except if the event is external
or the child node is shareable. 
What solution would you prefer ?

> Duplicate entries in the index
> ------------------------------
>
>                 Key: JCR-2357
>                 URL: https://issues.apache.org/jira/browse/JCR-2357
>             Project: Jackrabbit Content Repository
>          Issue Type: Bug
>          Components: indexing
>    Affects Versions: 1.5.0, 1.6.0, 2.0-alpha9
>            Reporter: Benjamin Papez
>            Priority: Critical
>
> We are very frequently having problems with queries, because the results are always duplicated.
The reason for it has been identified to be in the indexing, because when we get the problem,
we now look at the index with Luke, and we can see that documents with the same UUID are there
twice. This happens with Jackrabbit 1.5.0, 1.6.0 and also with 2.0-alpha10. 
> I am able to recreate the problem quite regular. We have some event listeners, which
themselves call the node.save or session.save method. I made a breakpoint in NodeIndexer.createDoc()
and stop when there is a specific name in the fullpath property. I can see that two listeners
we wrote ( FieldReferenceListener and RulesListener) lead to the NodeIndexer.createDoc() -
below are the stacktraces (based on Jackrabbit 2.0-alpha10)
>  NodeIndexer.createDoc() line: 247 
>  SearchIndex.createDocument(NodeState, NamespaceMappings, IndexFormatVersion) line: 1122

>  SearchIndex.updateNodes(Iterator<NodeId>, Iterator<NodeState>) line: 635

>  SearchManager.onEvent(EventIterator) line: 459 
>  EventConsumer.consumeEvents(EventStateCollection) line: 244 
>  ObservationDispatcher.dispatchEvents(EventStateCollection) line: 198 
>  EventStateCollection.dispatch() line: 455 
>  SharedItemStateManager$Update.end() line: 756 
>  SharedItemStateManager.update(ChangeLog, EventStateCollectionFactory) line: 1102 
>  XAItemStateManager(LocalItemStateManager).update(ChangeLog) line: 351 
>  XAItemStateManager.update(ChangeLog) line: 354 
>  XAItemStateManager(LocalItemStateManager).update() line: 326 
>  SessionItemStateManager.update() line: 326 
>  NodeImpl(ItemImpl).save() line: 1098 
>  FieldReferenceListener.updateFullPath(Node) line: 193 
>  FieldReferenceListener.onEvent(EventIterator) line: 101 
>  EventConsumer.consumeEvents(EventStateCollection) line: 244 
>  ObservationDispatcher.dispatchEvents(EventStateCollection) line: 198 
>  EventStateCollection.dispatch() line: 455 
>  SharedItemStateManager$Update.end() line: 756 
>  SharedItemStateManager.update(ChangeLog, EventStateCollectionFactory) line: 1102 
>  XAItemStateManager(LocalItemStateManager).update(ChangeLog) line: 351 
>  XAItemStateManager.update(ChangeLog) line: 354 
>  XAItemStateManager(LocalItemStateManager).update() line: 326 
>  SessionItemStateManager.update() line: 326 
>  NodeImpl(ItemImpl).save() line: 1098 
>  NodeIndexer.createDoc() line: 247 
>  SearchIndex.createDocument(NodeState, NamespaceMappings, IndexFormatVersion) line: 1122

>  SearchIndex.updateNodes(Iterator<NodeId>, Iterator<NodeState>) line: 635

>  SearchManager.onEvent(EventIterator) line: 459 
>  EventConsumer.consumeEvents(EventStateCollection) line: 244 
>  ObservationDispatcher.dispatchEvents(EventStateCollection) line: 198 
>  EventStateCollection.dispatch() line: 455 
>  SharedItemStateManager$Update.end() line: 756 
>  SharedItemStateManager.update(ChangeLog, EventStateCollectionFactory) line: 1102 
>  XAItemStateManager(LocalItemStateManager).update(ChangeLog) line: 351 
>  XAItemStateManager.update(ChangeLog) line: 354 
>  XAItemStateManager(LocalItemStateManager).update() line: 326 
>  SessionItemStateManager.update() line: 326 
>  NodeImpl(ItemImpl).save() line: 1098 
>  XASessionImpl(SessionImpl).save() line: 928 
>  RulesListener.onEvent(EventIterator) line: 376 
>  EventConsumer.consumeEvents(EventStateCollection) line: 244 
>  ObservationDispatcher.dispatchEvents(EventStateCollection) line: 198 
>  EventStateCollection.dispatch() line: 455 
>  SharedItemStateManager$Update.end() line: 756 
>  SharedItemStateManager.update(ChangeLog, EventStateCollectionFactory) line: 1102 
>  XAItemStateManager(LocalItemStateManager).update(ChangeLog) line: 351 
>  XAItemStateManager.update(ChangeLog) line: 354 
>  XAItemStateManager(LocalItemStateManager).update() line: 326 
>  SessionItemStateManager.update() line: 326 
>  NodeImpl(ItemImpl).save() line: 1098 
> Now when continuing to debug after the second NodeIndexer.createDoc() I see that the
node with the watched UUID is also passed within the remove collection into  SearchIndex.updateNodes(Iterator<NodeId>,
Iterator<NodeState>) , but when looking at the index with Luke before and after finishing
this SearchIndex.updateNodes, I see that the first document with the UUID was NOT deleted
and then we always have two documents with the same UUID.
> Now it would be interesting, why the deletion failed. When debugging DeleteNode.exceute
and stopping at the watched UUID, I come down the following methods:
> CommittableIndexReader(IndexReader).deleteDocuments(Term) line: 873	
> PersistentIndex(AbstractIndex).removeDocument(Term) line: 225	
> PersistentIndex.removeDocument(Term) line: 90	
> MultiIndex$DeleteNode.execute(MultiIndex) line: 1951	
> MultiIndex.executeAndLog(MultiIndex$Action) line: 1084	
> MultiIndex.update(Collection<NodeId>, Collection<Document>) line: 442	
> SearchIndex.updateNodes(Iterator<NodeId>, Iterator<NodeState>) line: 645

> What I can see in IndexReader.deleteDocuments(Term) is that docs.next() returns true,
but docs.doc() then returns 0 as the doc-id, which I think is the cause that the right doc-id
is not deleted. 
> Any clues what goes wrong here or hints what else I could watch for on debugging ?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message