ode-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "sijo cherian (JIRA)" <j...@apache.org>
Subject [jira] Created: (ODE-902) OutOfMemoryError / Heap usage issue under load test
Date Fri, 17 Dec 2010 20:00:10 GMT
OutOfMemoryError / Heap usage issue under load test
---------------------------------------------------

                 Key: ODE-902
                 URL: https://issues.apache.org/jira/browse/ODE-902
             Project: ODE
          Issue Type: Bug
          Components: BPEL Runtime
    Affects Versions: 1.3.4
         Environment: windowsXPprofessional or CentOS 4.8
            Reporter: sijo cherian


I am running one In-memory process, seeing OutOfMemory when under jmeter load test in ODE
1.3.4. Our bpel sequence has five webservice invokes (with very few database IO). The individual
webservice under same load test have no memory issues.

Heap profiling using JConsole:
Initial heap usage after test starts: ~200MB; peak usage upto 1Gb in under an hour (with constant
1 user load). In 2hrs the minimum heap 1GB (Garbage collection not able to reclaim any significant
memory). After that ODE ReadTimeout errors, followed by OutOfMemory error.

Using dehydration with these lenient config
ode-axis2.process.dehydration=true
# wait for 5 minutes instead of 20 minutes
ode-axis2.process.dehydration.maximum.age=300000
# allow not more than 200 processes to be in memory at once
ode-axis2.process.dehydration.maximum.count=200
# automatically discard any in-memory process instance after 4 minutes
ode-axis2.mex.inmem.ttl=240000

Load: One thread JMeter client with a delay(ConstantTimer) of 2sec.
Env: Dual core, 3 GB RAM, Tomcat 6.0.29, Java 6 (-Xmx1024 -XX:MaxPermSize=128m), ODE 1.3.4

On running a load spike (of about 500 requests), I see this heap behavior:
Initial heap usage 125mb. On sending 10 ODEprocess requests, heap peak usage 300 mb, then
back to ~125mb after GC (load test finished).
On sending 500 ODEprocess requests, heap peak usage 650 mb, then during all the inactivity
thereafter(tested for 12hrs), GC only brings back heap to 350mb. Looks like something is holding
on to the heap.
If I call only the individual webservices(instead of ODE process), peak heap usage is stable
at 200mb for 12hrs.

May be related to ODE-836

-- 
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