Shmul,
Ok, in that case. Currently, it is not possible to detect and log
that the JVM is
frozen and then not shut it down. You should be able to configure the
Wrapper so
that it never detects that the JVM is frozen however.
The Wrapper has the ability to detect a number of JVM failure
modes. The
freeze detection code works by sending pings to the JVM across a
socket. The
Wrapper then allows a configurable number of seconds for the JVM to
respond to
the ping. If that period ever times out then the Wrapper decides that
the JVM is
frozen and restarts it.
This timeout is configurable using the following property. A value
of 0 means
that the pings should never timeout and should be what you want.
wrapper.ping.timeout=0
If your app is hanging while it is starting up or shutting down then
other times
may need to be modified.
Your original message says that you had set the ping timeout to 0.
That was
not working for you? If not, try enabling debug output. I would need
that
information to help further.
Cheers,
Leif
Shmulik Regev wrote:
>>First of all 99% of the time. You do not want to be playing with the
>>timeouts. So start by removing those from your wrapper.conf file.
>>
>>I am making a wild guess here because you did not post much
>>information. But is your application timing out and shutting down
>>immediately after startup? This is usually caused because you
>>specified your application's main class directly in the wrapper.conf
>>file rather than using one of the helper classes like
>>WrapperSimpleApp. Make sure that you have read over the integration
>>section of the documentation. Most people make use of integration
>>method #1.
>>
>>
>
>Actually this is not the case for me. The application starts-up ok
>and the problem is surely not with the wrapper but with the
>application. What we see is that during stress testing (which we
>usually run at night, hence part of the problem), the jvm will not
>respond to pings and is shut down by the wrapper. The freeze of the
>jvm is probably caused by our own bug, however once the jvm is
>shutdown it is much harder for us to trace the cause of the problem.
>I would prefer to disable the shutting down of the jvm at least
>during development time to be able to get to the source of the
>application problem. It would actually be great if the wrapper would
>log the fact the the jvm doesn't respond, but will not terminate it.
>
>I'll also follow your advise regarding the timeout settings.
>
>Cheers,
>Shmul
>
>
|
