[infinispan-issues] [JBoss JIRA] Reopened: (ISPN-149) Infinispan transaction commit with many modifications

[Galder Zamarreno (JIRA)]

     [ https://jira.jboss.org/jira/browse/ISPN-149?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Galder Zamarreno reopened ISPN-149:
-----------------------------------

      Assignee: Galder Zamarreno  (was: Mircea Markus)


Reopening to implement suggested changes on IRC:

Aug 11 15:46:37 <galderz>	hi manik
Aug 11 15:47:05 <manik>	galderz hi
Aug 11 15:47:12 <galderz>	if we need to write a positive number bigger than a byte, what would you suggest? writing it as a short, assuming short is good enough, or use a unsigned int which has variable length?
Aug 11 15:48:58 <galderz>	dmlloyd too, thoughts on the above? mmarkus_, u might be interested in hearing other people's opinion
Aug 11 15:49:06 <galderz>	personally, i think i'd stick to a short
Aug 11 15:50:03 <manik>	galderz what is this for?
Aug 11 15:50:48 <galderz>	ReplicableCommandExternalizer, send the number of arguments
Aug 11 15:51:02 <galderz>	we're previously using a byte but that was not good enough, since max was 127 params
Aug 11 15:51:11 <galderz>	and in long transactions, this was been exceeded
Aug 11 15:51:31 <galderz>	https://jira.jboss.org/jira/browse/ISPN-149
Aug 11 15:58:04 <dmlloyd>	use both
Aug 11 15:58:17 <dmlloyd>	use a different ID byte if it's going to be > 255
Aug 11 15:58:23 <dmlloyd>	(also, use an unsigned byte or unsigned short)
Aug 11 15:58:43 <dmlloyd>	then it can be 1 byte in the majority of cases, or 2 if you need the extra range
Aug 11 16:08:19 <manik>	galderz yeah, otherwise a variable length int
Aug 11 16:09:05 <galderz>	dmlloyd, manik, ok
Aug 11 16:09:19 <galderz>	mmarkus_, do u want me to fix this accordingly?

Assigning it to myself.

> Infinispan transaction commit with many modifications
> -----------------------------------------------------
>
>                 Key: ISPN-149
>                 URL: https://jira.jboss.org/jira/browse/ISPN-149
>             Project: Infinispan
>          Issue Type: Bug
>    Affects Versions: 4.0.0.ALPHA6
>            Reporter: Mircea Markus
>            Assignee: Galder Zamarreno
>             Fix For: 4.0.0.BETA1
>
>
> Email from Lukasz Moren
> ----------------------------
> Hello,
> During adapting Infinispan into Hibernate Search I found that trying to commit transaction with many modifications (about 200 or more)
> on cache causes an exception - only if there is more than one member in the cluster. 
> This same problem appears when try to call endBatch(). 
> i.e.:
> begin transaction
> for ( int i = 0; i < 200; i++ )
>          cache.put(key, value);
> commit transaction
> I'm not sure if I do something wrong or it is Infinispan problem. Do you have some idea?  
> Below is configuration I use.
> 		
>                 GlobalConfiguration gc = GlobalConfiguration.getClusteredDefault();
> 		Configuration c = new Configuration();
> 		c.setInvocationBatchingEnabled( true );
> 		c.setCacheMode( Configuration.CacheMode.REPL_SYNC );
> 		c.setSyncReplTimeout( 30000 );
> 		c.setLockAcquisitionTimeout( 60000 );
> 		c.setUseLockStriping( false );
> 		CacheManager manager = new DefaultCacheManager( gc, c );
> 		cache = manager.getCache( "TestCache" );
> 		manager = new DefaultCacheManager( gc, c );
> 		cache2 = manager.getCache( "TestCache" );
> ------------------------

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