Administrative requests hang and default queue is gone

Question

Hi,
We have a recurring problem with our WebLogic 6.1 SP 1 server.
Periodically the server will hang and not respond to admin commands,
not
even a "java weblogic.Admin PING". It has to be stopped by killing
the OS process.
We've dumped the threads using SIGQUIT and examined them for
deadlocks, but something more striking turned up: all the "default
queue" threads were gone. In a healthy, live server there were
something like 60 threads, several of them from the "default queue",
but in our hung servers, these would be nonexistent.
From what I understand, the default queue contains the threads that do
most of the work, like servicing EJB requests, so it makes sense that
if they're not there, the server would not respond to anything. But I
don't see how they could die without there being any kind of
information about their death. There's nothing about it in the
weblogic server log, and nothing in stdout or stderr.
Has anybody experienced the same behavior? Does anyone know what
causes it?
Thanks,
Jim Doyle
[email protected]

Answer 1

This is surprising. Are you seeing any OutOfMemory errors or strain on the
server's resources ?
Is the server under load ?
Jim Doyle wrote:
Hi,
We have a recurring problem with our WebLogic 6.1 SP 1 server.
Periodically the server will hang and not respond to admin commands,
not
even a "java weblogic.Admin PING". It has to be stopped by killing
the OS process.
We've dumped the threads using SIGQUIT and examined them for
deadlocks, but something more striking turned up: all the "default
queue" threads were gone. In a healthy, live server there were
something like 60 threads, several of them from the "default queue",
but in our hung servers, these would be nonexistent.
From what I understand, the default queue contains the threads that do
most of the work, like servicing EJB requests, so it makes sense that
if they're not there, the server would not respond to anything. But I
don't see how they could die without there being any kind of
information about their death. There's nothing about it in the
weblogic server log, and nothing in stdout or stderr.
Has anybody experienced the same behavior? Does anyone know what
causes it?
Thanks,
Jim Doyle
[email protected]