Jorn, I will try to find a source of this leakage today. There is a per CAS
statistic cached in a Map in the UIMA AS service and most likely this Map is
not being cleaned up every time a CAS is returned to a client. I will create
JIRA for this once I confirm the problem.
Thanks
Jerry C
On Thu, Apr 15, 2010 at 10:22 AM, Jörn Kottmann <kottmann@gmail.com> wrote:
> Hi everyone,
>
> after running a UIMA-AS node for 6 days I noticed
> that there are around 400K instances of
> org.apache.uima.aae.jmx.ServicePerformance
> objects consuming around 80 MB of memory.
>
> In our production system the UIMA-AS worker nodes
> fail after a few weeks with an OutOfMemoryException
> and I would like to know if the huge amount of ServicePerformance
> objects might be linked to that failure.
>
> Jörn
>
|