Re: [Wrapper-user] Timeout on heavy server startup

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

Ori,
Version 3.1.0 fixed a problem where the first ping timeout was effectively
hard coded to 30 seconds even though you set the wrapper.ping.timeout
property
to a larger value. From your output, I think this might be what you are
running
into.

Could you try using 3.1.0 and let me know how it works. If it is still
failing,
then I would need you to set the wrapper.debug=true property and post the
resulting wrapper.log file so I can tell you exactly how to resolve it.

Cheers,
Leif

Ori Argov wrote:

> Hi Lief,
>
> I'm using version 3.0.2 of the wrapper and it generally runs great.
>
> I'm using the JSW to start a heavy weblogic server which takes quite a
> while to start.
>
> I've implemented the necessary interface to control the startup of the
> NT service and
>
> Ask for waitHints as the server goes up. (Sometime it even maxes out
> on the windows timeout for service startup, but I've seen Oracle do
> that as well, so I guess this cannot be helped by the wrapper which
> uses the NT Service Manager like every other service).
>
> The issue I'm seeing is that rarely (not a specific machine or OS) the
> wrapper decides to timeout the server for no apparent reason.
>
> The startup timeout is set to 1200 seconds, and the ping timeout to
> 600 (yes, I know it's bad - but we are using the wrapper mainly for
> its service wrapping capabilities and not for JVM hanging detections)
> but the server timeouts after about 2 minutes.
>
> The waitHint is 15 seconds each time.
>
> Cpu.timeout is 10 sec.
>
> The error is this:
>
> ERROR | wrapper | 2004/09/26 09:08:48 | Startup failed: Timed out
> waiting for signal from JVM.
>
> ERROR | wrapper | 2004/09/26 09:08:48 | Java Virtual Machine did not
> exit on request, terminated
>
> STATUS | wrapper | 2004/09/26 09:08:54 | Launching a JVM...
>
> I have looked at the code and saw 2 places where this error is being
> issued: (in wrapper.c)
>
> Line 1592: this looks like in a JVM launch state, so we passed that
> already.
>
> Line 1638: this looks like the right one since we are in startup state.
>
> What is suspicious for me is that the same variable
> wrapperData->jStateTimeout is used to keep track of all timeout events.
>
> This field is being updated both on pings and waitHints. I know that
> it is always incremented from time(NULL) but could
>
> It be that the ping (which is smaller) overrides the startup waitHint
> on this variable and causes a timeout?
>
> I have not analyzed the code completely so forgive me if I've just
> said something silly.
>
> Any ideas on how to pinpoint this problem? Which prints should I add
> that will help me analyze the error next time I see it?
>
> Thanks,
>
> Ori
>