RE: Listener stops accepting client requests/connections
- From: "Jeremiah Wilton" <jwilton@xxxxxxxxxxxxx>
- To: <EVAD@xxxxxxxxxx>, <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 12 Nov 2004 10:46:04 -0800
Denham,
If the stuck listener is still visible in the processes tab of the task
manager, you could get a crash dump of the stuck listener with the Dr.
Watson debugger.
Also, is the listener responding to "lsnrctl status" or "lsnrctl services"?
If so it is not really stuck and the problem may be harder to determine.
If it is genuinely unresponsive to resuests of any kind, first determine if
it is stuck or spinning. In the processes tab of the task manager, sort by
CPU descending. If the stuck listener registers much CPU at all it is
spinning. If it registers zero, it is stuck.
You'll need the process ID so you have to make that get displayed in the
task manager under View -> Select Columns
The drwtsn32.log is usually pretty full on most people's systems so rename
the file C:\Documents and Settings\All Users\Application Data\Microsoft\Dr
Watson\drwtsn32.log to something else. That way you'll get a fresh log file
when you obtain the listener crash dump.
Crash the stuck listener with Dr. Watson like so:
Start -> Run -> drwtsn32 -p <process ID>
Where <process ID> is the ID of the stuck listener.
The resulting drwtsn32.log will contain a call stack trace and other useful
information for a *competent* Oracle support analyst. Hint: if no analyst at
Support has suggested a course of action similar to this, the you have yet
to get a competent analyst.
This is much a much more cumbersome diagnostic task on Windows than on any
type of Unix. Those with more Windows knowledge than I may know of a better
debugger than Dr. Watson that might actually allow you to attach, get a
stack trace, then detach without killing the listener. Also, someone may
know of a system call tracing utility for Windows that performs similarly to
strace/truss/tusc. Output from such a utility attached to the stuck listener
would be very useful.
--
Jeremiah Wilton
Independent Oracle Professional
www.speakeasy.net/~jwilton
-----Original Message-----
We have two servers a 9.2.0.3 and 8.1.7.4.1 both on Windows 2000 Servers
Both these servers' listeners have started an odd behavior, for no
apparent reason they stop receiving/accepting client access.
However if you are connected you remain connected. There is also no
specific time base for this behavior. Once the Listener services have
been bounced everything appears normal for a day or so, then it happens
again.
Strangely if we have had to stop and start the server because of
parameter changes/anything the Listener is fine until a day or two
thereafter.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
