[jboss-jira] [JBoss JIRA] Commented: (JBCACHE-1123) Non-streaming state transfer failures not properly handled

[Brian Stansberry (JIRA)]

    [ http://jira.jboss.com/jira/browse/JBCACHE-1123?page=comments#action_12368079 ] 
            
Brian Stansberry commented on JBCACHE-1123:
-------------------------------------------

Yep, I figured something like that.

> Non-streaming state transfer failures not properly handled
> ----------------------------------------------------------
>
>                 Key: JBCACHE-1123
>                 URL: http://jira.jboss.com/jira/browse/JBCACHE-1123
>             Project: JBoss Cache
>          Issue Type: Bug
>      Security Level: Public(Everyone can see) 
>          Components: Clustering
>    Affects Versions: 2.0.0.CR3
>            Reporter: Brian Stansberry
>         Assigned To: Vladimir Blagojevic
>             Fix For: 2.0.0.CR4
>
>
> I'm noticing spurious WARN messages in the logs from JGroups when caches do partial state transfers and the cache being requested doesn't have the target node available:
> WARN  [org.jgroups.protocols.pbcast.STATE_TRANSFER] state received from 192.168.1.164:4655 is null, will return null state to application
> Looking at the JBC state transfer code, particularly StateTransferManager.getState(ObjectOutputStream out, Fqn fqn, long timeout, boolean force, boolean suppressErrors), the apparent intent is for a boolean 'false' and a CacheException to be written to the stream, serialized, and returned to the requestor. The CacheException is also thrown.
> This is breaking in CacheImpl.MessageListenerAdaptor.getState() and getState(String):
>          try
>          {
>             ExposedByteArrayOutputStream baos = new ExposedByteArrayOutputStream(16 * 1024);
>             out = new MarshalledValueOutputStream(baos);
>             getStateTransferManager().getState(out, Fqn.fromString(sourceRoot),
>                     configuration.getStateRetrievalTimeout(), true, true);
>             result = baos.getRawBuffer();
>          }
>          catch (Throwable t)
>          {
>             stateProducingFailed(t);
>          }
>          finally
>          {
>             Util.close(out);
>          }
>          return result;
> The CacheException is thrown from StateTransferManager.getState().  Therefore "result = baos.getRawBuffer();" is never invoked.  Therefore the returned state is null, leading to the WARN on the recipient.  This seems contrary to the intent to propagate the boolean + exception to the recipient.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira