Re: [Wrapper-user] How to disable wrapper stopping, when killing JVM ?

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

Martin,
I just retested this and it is working correctly for me. Most likely,
this is a misunderstanding about when it is used. The wrapper's
sh script by default will only use the ignore signals setting when it
is being run as a daemon process. Ie when "start" is used, not
"console".

The thinking here is that daemon processes are the ones that you
really want to be protecting from stray signals. "console" mode is
mainly used for development as the wrapper would be killed when the
shell was closed by default anyway.

Let me know your use case as it may require some changes.

If you want to ignore signals in console mode as well, you still need
to set the IGNORE_SIGNALS flag in the sh script to put the sh script
into that mode, but in your wrapper.conf file, add the following
property:
wrapper.ignore_signals=TRUE

When IGNORE_SIGNALS is set, the sh script also causes the wrapper to
create an anchor file. It then uses that to tell the wrapper to
shut down. When IGNORE_SIGNALS is not set, the sh script attempts
to shut down the wrapper by sending a TERM signal.

Cheers,
Leif

Martin Gazak wrote:
>  Good morning,
>  I am using Java Service Wrapper under Linux and I am facing following
>  problem:
>  When I kill the JVM process (not wrapper, but JVM) with signal "kill
>  -9", wrapper logs:
>
>  ERROR  | wrapper  | 2007/02/21 08:36:48 | JVM exited unexpectedly.
>  STATUS | wrapper  | 2007/02/21 08:36:53 | Launching a JVM...
>
>  and restarts the JVM which is what I want.
>
>  When I kill the JVM process (not wrapper, but JVM) with just "kill",
>  wrapper logs:
>
>  STATUS | wrapper  | 2007/02/21 08:32:25 | <-- Wrapper Stopped
>
>  and stops without restarting JVM.
>  Is there any way how to configure wrapper (or to change application ?)
>  to achieve, that wrapper restarts JVM no matter what signal JVM received ?
>
>  I played with wrapper.ignore_signal property, but it seems to work on
>  wrapper, not on the spawned JVM and therefore it did not help me.
>
>  Thanks.
>
>  With best regards
>
>  Martin Gazak
>  MicroStep-MIS
>