gobblin-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vicky Kak <vicky....@gmail.com>
Subject Re: Possible leak due to Metrics.
Date Fri, 10 Nov 2017 13:50:22 GMT
Hi Guys,

The issue seems to sorted by increasing the swap space. I too multiple
dumps and observed the leaks from the metrics code is not crossing 125 MB,
it does not seem to be leak.
https://www.build-business-websites.co.uk/java-memory-allocation-failures-due-to-insufficient-swap-space/

Thanks,
Vicky

On Thu, Nov 9, 2017 at 7:24 PM, Vicky Kak <vicky.kak@gmail.com> wrote:

> Hi Guys,
>
> We have got the following use case
>
> 1) The Custom server implemented in Gobblin which acts as a rest interface
> and launches the gobblin Job.
>
> 2) During the stress test the same job is being executed lots of times and
> I see the JVM crashing at time. I took the heap dump and analysed it using
> Eclipse MAT,
> the reports suggested the Metrics related classes causing it. Since we are
> launching same Jobs too many times in the JVM so it seems for each
> invocation the metrics
> related classes are getting created.Please find the attachment for the
> reference.
> I would be moving to the GAAS in some time and I do anticipate the same
> might be happening there too.
>
> Can we provide a way to clean the metrics once the Job is done or after
> some time?
>
> In our case our service will be running for ever so we will eventually hit
> the JVM heap peak, the same could be happening with the GAAS.
>
> Please let me know is there is any other options to get rid of the this
> issue.
>
> Thanks,
> Vicky
>

Mime
View raw message