|
From: Leif M. <le...@ta...> - 2007-10-30 00:40:04
|
Marcin,
Sorry, the Wrapper does not do anything fancy with class loaders or
anything else as that would have consequences for other applications.
It looks like Geronimo is not going to work as is. It should be very
easy to modify it so it would work with either method 1 or method 2
smoothly. Some applications even go as far as to support method 3
directly.
Cheers,
Leif
Marcin Waldowski wrote:
> Leif Mortenson wrote:
>
>> Marcin,
>> Without using the wrapper, many applications launch a second JVM
>> instance simply to trigger the shutdown of the first. This is usually
>> done by deleting an anchor file or sending a command via a socket.
>> The wrapper makes this much lighter by running this shutdown
>> code in the same JVM.
>>
>
> Yes, and this is OK! :)
>
>
>> Are you having problems running the shudown class in the same
>> JVM?
>>
>
> Yes, in case of Geronimo it result with IllegalStateException.
>
> In "start" class we have execution of:
> KernelRegistry.registerKernel("MainBootstrapper");
> and in "stop" class we have execution of the same:
> KernelRegistry.registerKernel("MainBootstrapper");
> which in "stop" class result with
> java.lang.IllegalStateException: A kernel is already running this
> kernel
>
> You can see discusion about it here:
> http://www.mail-archive.com/us...@ge.../msg07665.html
> Result of discussion is that it is *intended* to run shutdown class in a
> different vm. It means that second integration method can not be used
> (only first can be used) for Geronimo.
>
> Of course the best solution is to modify Geronimo shutdown class. But
> first I decided to ask You if it is posible to do some trick to make it
> works in second integration method ;)
>
> I'm *not* trying to tell you that it's worth to implement calling "stop"
> class in different vm in JSW. Definitely I think that this is *not*
> worth for this single case :) I'm just asking for some trick which meybe
> is implemented already ;)
>
> Regards, Marcin
|