|
From: Leif M. <le...@ta...> - 2004-03-18 07:56:55
|
Stefan, I was able to reproduce this. It is actually a problem on both Windows and UNIX, but is much more difficult to get the timing right to encounter it on Windows. The problem is that the Wrapper is receiving a SIGTERM after it has launched the JVM process but before the the JVM has come up far enough to have opened up its backend socket to the Wrapper. This means that the Wrapper is unable to tell the JVM to stop. That said, the Wrapper does go into shutdown mode, so as soon as the wrapper.shudown.timeout expires the Wrapper will kill the JVM. Unless you have extended that timeout, this should have been happening. I did fix the root problem however so the JVM will now be told to stop as soon as it opens up the backend socket. This should guarantee that the JVM shuts down cleanly within a few seconds rather than being shutdown forcibly after 30 seconds. The fix is already in CVS and will be in version 3.1.0 Cheers, Leif Pauwels Stefan wrote: >Hi, > >I do the following: > >script.sh start >and then immediately followed by >script.sh stop > >After this I see that the application is still running, >but the PID file is removed. > >The problem is that calling the constructor in my >application, takes a long time >because it needs to do a lot of initialisations. > >If I do: >script.sh stop, before the constructor has finished, >this stop is ignored. > >Can someone help me ? > >Thanks, >Stefan > > |