Re: [Wrapper-user] Advice on stopping wrapper from triggering JVM restarts

[Leif M. <lei...@ta...>]

Casey,
If you are able to reproduce this, can you set wrapper.debug=true and
repeat?  This log output will shed more light on exactly what is
happening.  Depending on the platform, we will also be able to see the pid
of the source of the kill signal.

Cheers,
Leif

On Wed, Dec 3, 2014 at 12:39 AM, Casey Jordan <cas...@jo...>
wrote:

> Hi Leif,
>
> Thanks for the feedback. This is quite a mystery then because I am 100%
> sure that the process wasn't killed manually or by any other systems we
> have running. (This is just a standard CentOS 7 setup)
>
> At about the same time that this log appeared I was running a very heavy
> build process in a separate JVM instance.
>
> It seems more than just coincidental that this happened at the same time.
> I assume that if the JVM crashed it wouldn't have received a SIG 9? Can you
> think of any scenario where this makes sense? Perhaps I am missing some
> knowledge of java architecture here that is making this mysterious.
>
> Thanks
>
>
> On Tue, Dec 2, 2014 at 4:27 AM, Leif Mortenson <
> lei...@ta...> wrote:
>
>> Casey,
>> The configuration you specified will disable all pings and cpu warnings
>> and should do what you want.
>> The Wrapper should never timeout  in this case.  There are other things
>> such as a JVM crash would would still result in a restart after the fact.
>>
>> The log that you send shows that the JVM will killed with a kill -9.  Not
>> sure if that was a test on your part, but if the source of the signal has
>> permission to kill the JVM's process then there is nothing we can do to
>> prevent that.
>>
>> Cheers,
>> Leif
>>
>> On Tue, Dec 2, 2014 at 12:16 PM, Casey Jordan <cas...@jo...>
>> wrote:
>>
>>> Hi all,
>>>
>>> I am sure this is a common question, but I am having trouble extracting
>>> details I need from the documentation. From time to time we get a situation
>>> where our JVM gets killed and we see something like this in the logs:
>>>
>>> INFO   | wrapper  | 2014/12/02 02:23:18 | Wrapper Process has not
>>> received any CPU time for 1 seconds.  Extending timeouts.
>>> STATUS | wrapper  | 2014/12/02 02:23:37 | JVM received a signal SIGKILL
>>> (9).
>>> STATUS | wrapper  | 2014/12/02 02:23:37 | JVM process is gone.
>>> ERROR  | wrapper  | 2014/12/02 02:23:37 | JVM exited unexpectedly.
>>> STATUS | wrapper  | 2014/12/02 02:23:38 | Automatic JVM Restarts
>>> disabled.  Shutting down.
>>> STATUS | wrapper  | 2014/12/02 02:23:38 | <-- Wrapper Stopped
>>>
>>>
>>> This always happens in the case where some other process is eating up
>>> all the cpu for more than a few seconds.
>>>
>>> It's crucial for us that the wrapper never, ever restarts the JVM as
>>> this is something we always want to look into manually.
>>>
>>> So my question is, what is the best way to make sure this doesn't happen?
>>>
>>> I am using community version 3.5.17, and have the following values in
>>> our wrapper.conf:
>>>
>>>    #********************************************************************
>>>    # Timeouts
>>>    #********************************************************************
>>>    wrapper.ping.timeout=0
>>>    wrapper.ping.timeout.action=DEBUG
>>>    wrapper.cpu.timeout=0
>>>    wrapper.startup.timeout=300
>>>
>>> Any help is much appreciated.
>>>
>>> Thanks!
>>>
>>>