Re: Cold backup blues (UNIX)...

From: nchintala <nchintala_at_worldnet.att.net>
Date: 1996/08/30
Message-ID: <3227A055.3932_at_worldnet.att.net>#1/1


Simon wrote:
>
> irvin_at_lmsc.lockheed.com (Tim Irvin) wrote:
>
> >Sometimes, "shutdown immediate" doesn't seem to work--Oracle hangs trying
> >to shut down. The datafiles get backed up--I'm not sure how much I trust
> >their integrity in this state--and when that happens, the startup fails
> >and the backup script is still going...
 

> >Some don't recommend it, but if it's a recurring problem, you could
> >replace a shutdown immediate with a shutdown abort, startup and shutdown
> >normal (or immediate).
>
> Tim, I get exactly the same problem - in fact, we had it so badly (ie
> 90% of the time that we tried to do a cold backup), that we have now
> replaced the "clean" method of shutdown immediate with the shutdown
> abort, startup, and normal shutdown approach.
>
> I don't like it, but it seems to be the only way that we can bring the
> database down to do the backup, and guarantee that it is going to
> work.
>
> We have... HP-UX 9.04, Oracle 6.0.37.3.1
>
> Is it an HP-UX issue, do you think?

I don't think so. We encountered the same problem on Solaris 2.5 with Oracle 7.2.3. Shutdown immediate hangs forever with a message in the alert*.log file "waiting for users to disconnect". There were no users connected to Oracle (I am sure of it because I was the only one logged in at that time).

Many of you have suggested the following

  1. shutdown abort.
  2. startup normal
  3. shutdown normal

Looks like this works for most of you.

Here is my question.

If you do a shutdown abort Oracle does not grauntee that it will clean the ipc resources ( I got this information from Oracle Support ). So the next time you start the instance you will get the error "Unable to complete internal login, warm start the instance". The oracle background process hand around sometimes. As far as I know the only
way to overcome this problem is to

  1. Use the FORCE otion during start (sometimes this does not work).
  2. clean the ipc resources by the ipcrm command manually.

What if there are multiple instances on the same machine? How does one identify the ipc resources related to one particular instance?

Thanks for any input regarding this problem. Please email also.

Naren Chintala
nchintala_at_worldnet.att.net
naren_at_mink.att.com

> --
> Simon Holt -- System Designer/DBA
> ** My opinions are not nescessarily those of my employers **
> Britannia Zinc Ltd. Bristol, UK. (+44) 0117 982 3646
Received on Fri Aug 30 1996 - 00:00:00 CEST

Original text of this message