RE: DATABASE HANGS WITH SHUTDOWN IMMEDIATE

From: Powell, Mark D <mark.powell_at_eds.com>
Date: Fri, 28 Apr 2006 09:20:07 -0400
Message-ID: <5A14AF34CFF8AD44A44891F7C9FF4105072D1120@usahm236.amer.corp.eds.com>

We had multiple hanging problems on 9.2.0.4 RAC running on AIX 5.2. 9.2.0.5 and later 9.2.0.6 were better though we have had a hang or two. I can only remember the details for two of the three separate bugs that we hit and which support identified as the cause of our hung database. The first seems unlikely since you are using one-node RAC. In a RAC environment if too many sessions attempt to log on too concurrently the database can develop a problem trying to access the sequence generator behind v$session.audsid. Just increase the cache size from 20 to 1,000. (Support recommended 10k but that seems a little like overkill)  

The second was identified as not being a bug but a deadlock occurred on the rdbms base tables. Oracle is apparently unable to identify this condition like it can for user table access so it does not kill one of the sessions and roll it back automatically. However, you can however see the problem in the hang analyze dump.  

We hit at least one other identified bug but I do not remember the details at all.  

Next time before issuing any shutdown command run a hang analyze which will give you something to look at. If you cannot spot the problem in the trace then you upload the trace to support to give them something to work with.  

HTH -- Mark D Powell --  

---

        From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Ahmed Ullah

	Sent: Thursday, April 27, 2006 6:36 PM
	To: oracle-l_at_freelists.org
	Subject: DATABASE HANGS WITH SHUTDOWN IMMEDIATE 
	
	
	
	Hello all:
	 
	We have Oracle 9.2.0.5 database ( single Note RAC ). There were

some heavy batch jobs running which cuased the db hang ( only sys connection was working ).          

        We tried to bring down the database after cleaning all sessions. It did not go down with SHUTDOWN IMMEDIATE. We had to abort it. I tried to bring down the DB three times with SHUTDOWN IMMEDIATE it never went down.          

        SMON was trying to cleanup the used extents ( Metalink Note - 1076161.6 ) as we could see in the alert log file - Waiting for smon to disable tx recovery.          

        SQL> select count(block#) from fet$;          

	COUNT(BLOCK#)


	-------------

	           38
	 
	fet$ - free extents
	 
	SQL> select count(block#) from uet$;
	 
	COUNT(BLOCK#)


	-------------

	        52905
	 
	uet$ - used extents
	 
	We can share the alert log and system level 10 trace files if

anyone wants. These traces were generated when db hung.          

        Appreciate your help.          

        Thanks,          

        -ahmed          

---

	"If we knew what it was we were doing, it wouldn't
	be called research, would it?" --Albert Einstein
	 

	
________________________________

	New Yahoo! Messenger with Voice. Call regular phones from your

PC
<http://us.rd.yahoo.com/mail_us/taglines/postman5/*http://us.rd.yahoo.co m/evt=39666/*http://messenger.yahoo.com> and save big.

--
http://www.freelists.org/webpage/oracle-l

Received on Fri Apr 28 2006 - 08:20:07 CDT