agent cannot connect to the server. java.lang.ClassNotFoundException: org.hibernate.collection.PersistentMap

John Mazzitelli mazz at redhat.com
Fri Jun 25 15:56:42 UTC 2010


> Is there a way to fix it? Or do I have to un-inventory the platform and
> add it again? (which I obviously wouldnt want to do as I will loose all
> the config stuff)

I've not seen a case where the agent gets so totally corrupted that a 
restart with --cleanconfig doesn't fix the problem. Usually, if you do a 
--cleanconfig restart, the agent starts up clean (you even have to setup 
the original configuration by answering the questions) and it will pull 
down the full inventory from the server. Is that not happening? Worst 
case, yes, shutdown the agent, uninventory the platform, and restart the 
agent with --cleanconfig. But again, its rare that that would have to 
get done - not seen a recent case where that would be necessary.

> BTW what can be the reasons of memory issues? Are you aware of any
> memory leaks?

I'm at this very moment running JProfiler on the agent as its monitoring 
two JBossAS 5 instances and the RHQ Server (which is a JBossAS 4 
instance). I'm not seeing anything abnormal right now. I'm profiling 
some more so we'll see what happens.

 > Will it help if I configure JVM to automatically create a
> dump in case of OOME and send it to devs to investigate? What are the
> recommendations re heap size? (eg number of metrics -> heap max size)

That would be helpful. I usually like to see what the thread dump looks 
like - seeing what threads are running at the time usually can help 
pinpoint the problem. As for heap size, it really depends on what you 
are monitoring - if you are monitoring a single JBossAS instance , the 
heap requirements are much less than if you are monitoring 20 instances 
of JbossAS. When in doubt, and if you have the memory, make the heap 
512m at least.  That's just a guess, but I'm thinking that should be 
enough in the more general case.

> Sorry about so many questions. Im just a newbie in RHQ, had not very
> successful experience with Hyperic and its way to split open source and
> commercial functionality, and very excited with the RHQ approach
> allowing to do a lot of stuff programmatically or with couple of mouse
> clicks (eg dynagroups) which makes it absolutely perfect to manage
> hundreds of our servers!

Which is one of the main reasons Dynagroups was implemented. How many 
servers are you managing? How many resources are each agent managing 
(e.g. how many JBossAS servers do you have running on a single box that 
an agent is monitoring?)


More information about the rhq-users mailing list