Bug 534344 (RHQ-114)

double check that we aren't doing what Steve asked about:

http://opensource.atlassian.com/projects/hibernate/browse/HHH-3006#action_29503

we are using EJB3/JPA to handle our hibernate session management so I'm not sure if its possible that we could do this, but we are definitely processing content reports concurrently on differernt threads (but each thread calls into the SLSB seperately, so they all should get their own session).

(4:22:28 PM) ccrouch: thanks, so do we care about RHQ-114? are we seeing any sideeffects?
(4:25:29 PM) mazz: I only saw RHQ-114 once
(4:25:39 PM) mazz: but never found out why and I did not do anything explicitly to fix it
(4:25:46 PM) mazz: so I suspect its possible to happen again
(4:26:07 PM) ccrouch: right, did it actually break anything?
(4:27:15 PM) mazz: well, the logs spun out of control and never stopped... got MBs of logs due to that infinite loop
(4:29:07 PM) ccrouch: I guesss we could turn down the logging, if that was the only problem
(4:29:47 PM) mazz: well, that means we have a spinning thread

I think I was too hasty on making these Blockers, they're not good but the side effect so far appear minimal, or we eventually recover.

This isnt critical for the 1.0 release given the reduction in amount of content being discovered

putting on 1.1 table for triage, jdobies please review and determine whether 2.0 is a better timeframe.

now that we have the newly created 1.2 release, we'll get this off of 1.1's plate since there likely won't be enough time to do another full round of performance testing.

looking at that HHH JIRA, it looks like Steve added some code to fix the issue (or at least workaround it - since the cause might be due to txs timing out).  Unsure if we want to upgrade our hibernate but that's what it would take to address this.

this happens not just with large content reports, but I'm seeing this when just starting up new agents (5 agent spawn VMs, 5 agents per VM).

we need to examine the possibility of upgrading hibernate - this problem is too severe to ignore.

This problem causes 230K events to be inserted in the database per hour. Not good.

FYI: our current version of hibernate is from r14210 of hibernate branch "branch_3_2 " - which I think can now be found here "http://anonsvn.jboss.org/repos/hibernate/core/branches/Branch_3_2/" - with this patch applied:

Index: EntityModeToTuplizerMapping.java
===================================================================
--- EntityModeToTuplizerMapping.java	(revision 14201)
+++ EntityModeToTuplizerMapping.java	(working copy)
@@ -18,7 +18,7 @@
 public abstract class EntityModeToTuplizerMapping implements Serializable {
 
 	// map of EntityMode -> Tuplizer
-	private final Map tuplizers = Collections.synchronizedMap( new SequencedHashMap() );
+	private final Map tuplizers = new org.hibernate.util.FastHashMap();
 
 	protected void addTuplizer(EntityMode entityMode, Tuplizer tuplizer) {
 		tuplizers.put( entityMode, tuplizer );

the patch in the previous comment doesn't have a full path to the patched file - so to be clear, it patches this file: http://anonsvn.jboss.org/repos/hibernate/core/branches/Branch_3_2/src/org/hibernate/tuple/EntityModeToTuplizerMapping.java

"John Mazzitelli - 14/Nov/08 04:28 PM
FYI: our current version of hibernate is from r14210 of hibernate branch "branch_3_2 " "

I don't believe is correct. I think it should say ".. r14201..". This came from a typo I made in https://jira.jboss.org/jira/browse/JBNADM-2822. For confirmation look at the patch...

Index: EntityModeToTuplizerMapping.java
===================================================================
--- EntityModeToTuplizerMapping.java (revision 14201) 

Patched hibernate build instructions:
svn co -r 14201 http://anonsvn.jboss.org/repos/hibernate/core/branches/Branch_3_2/

Apply the attached patches to src/org/hibernate/jdbc/AbstractBatcher.java and src/org/hibernate/tuple/EntityModeToTuplizerMapping.java

Then build the hibernate jar using ./build and then \Branch_3_2\build\hibernate3.jar is the jar you want

hibernate3.jar uploaded on Nov17 has md5 of 6d5a50a98a0a62f73daf795da202307e and is built from r14201 of hibernate 3_2 branch plus the patches to AbstractBatcher.java and EntityModeToTuplizerMapping.java

----- "Ian Springer" <ian.springer> wrote:

> It's up:
> 
> http://repository.jboss.com/maven2/hibernate/hibernate3/3.2.r14201-2/

still need to apply to the 1.1.2 cp branch when we deem it appropriate - see this issue's subtask

QA Closing, this issue has not recurred in the performance environment for some time.

This bug was previously known as http://jira.rhq-project.org/browse/RHQ-114
Imported an attachment (id=368500)
Imported an attachment (id=368501)
Imported an attachment (id=368502)
This bug is related to RHQ-1122