RE: Database/system Crashing

[Richard Ji]

Title: RE: Database/system Crashing

Sorry for getting back to you late on this.  In our case, the server crashes each time with CPU panic message

in /var/adm/messages.  We used adb to analyze the core dump and saw in each instance of the crash, NFS

lead to it.  It was trying to free memory twice which caused the panic.  And talking to Sun confirmed that there

was a bug in NFS.  If you are interested in it I can send you the patch number.

-----Original Message-----
From: Webber Valerie H [mailto:[EMAIL PROTECTED]]
Sent: Monday, January 06, 2003 5:26 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Database/system Crashing

Richard,

 

Yes this is a SUN platform. One SA now believes it is independent of the database since it happens when the database isn't running. I moved the background dest files to the other disk (other than the root disk.) Normally it would have crashed by now after starting the database and using Designer but so far so good.

 

There are no error messages in /var/adm/messages. There's no core dump file either. Its like the system gets corrupt before it can write to the /var/adm/messages file.

 

What were the symptoms of your Sun NFS related crashed and how did you diagnois the problem and what was the solution (if it were that simple)

 

Val

-----Original Message-----
From: Richard Ji [mailto:[EMAIL PROTECTED]]
Sent: Monday, January 06, 2003 4:19 PM
To: Multiple recipients of list ORACLE-L
Subject: RE: Database/system Crashing

Val,

 

Sorry I missed the previous messages.  Was this a Sun platform?  Did the system crash with a CPU panic

in /var/adm/messages?  We resovled a Sun NFS related crashes a couple of months ago.

 

Richard Ji

-----Original Message-----
From: Webber Valerie H [mailto:[EMAIL PROTECTED]]
Sent: Monday, January 06, 2003 11:40 AM
To: Multiple recipients of list ORACLE-L
Subject: RE: Database/system Crashing

Yes there are NFS mounts involved. What you said about the OS locks on the audit directory makes a lot of sense. My SA's are back to thinking it's a OS problem because it crashed again with the database shut down.

The odd thing is that there is nothing written to the Oracle alert log file nor are there any entries in the trace files. But when the system is rebooted and I bring the db back up, Oracle knows it previously crashed and recovers itself. That's in the alert log file. Its like the system is losing its pointers or something. I suggested reinstalling the OS and Oracle then put my database back and see if that helps. Are there huge risks with this scenario?

Another odd thing that the SA's can't figure out is there are no entries in the message file nor can they get a dump file to determine why the system crashed. There is nothing. It crashed over the weekend with no activity and they got some sort of i-nodes error.

Thanks for all your replies. Any ideas are helpful and I will relay them to our SA's...

Val

-----Original Message-----
From: Stephen Lee [mailto:[EMAIL PROTECTED]]
Sent: Friday, January 03, 2003 10:09 AM
To: Multiple recipients of list ORACLE-L
Subject: RE: Database/system Crashing

I wonder if a file lock is being left in place when the instance crashes,
and the OS does not clear the lock until a reboot.  I would think the OS
should clear this without a reboot, but stranger things have been seen with
OS's ... even Unix.  This doesn't explain why the instance crashes.  I
wonder if fuser would show anything.  Are there any NFS mounts involved?

-----Original Message-----
Yes, you're correct and it can write the file to $ORACLE_HOME/rdbms/audit
once the system is rebooted. Its just that when the database crashes, it
can't write to that location until its rebooted.
Is it possible that I need to beef up my init.ora parameters?
--
Please see the official ORACLE-L FAQ: http://www.orafaq.net
--
Author: Stephen Lee
  INET: [EMAIL PROTECTED]

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).