Re: shutdown abort vs. immediate

A shutdown immediate would be the option I would go with here. There would be really no need to script a shutdown immediate, then a shutdown abort if the immediate does not work. Shutdown immediate will shut the database down and leave it in a consistent state, where abort will not.

Regarding doing an abort because of a hardware failure and corruption, I don't buy the argument to do an abort to get off the failed hardware. The only time I see corruption happening is if you've setup the system with non-raid filesytems, plus, I hardly would believed that you're setting a clustered system and are not going to have any kind of Raid protection.

Regarding the redo log corruption, if you've set up multiplexed redo logs, then this should not be an issue at all.

I have experience with IBM HACMP & HAGEO on p660 6h1'a and have never run into such a issue with shutdown's. My cluster startup and failover is scripted for shutdown immediates with my cluster filesystems mirrored and have yet to experience a problem with Oracle not shutting down.

HTH,
Pete's

The veiws I may have expressed here are solely my own and not that of my employer.  

jon_e_price2003_at_yahoo.com (Jon) wrote in message news:<b81828cd.0306151710.6ddc9583_at_posting.google.com>...
> We have a Veritas clustered pair of servers running an Oracle based
> application.
> The application only runs on one system at any given time.
>
> Here's the issue:
>
> If the primary system has a hardware error that initiates a Veritas
> system failover, what are the advantages and disadvantages of using
> shutdown immediate versus shutdown abort to shutdown Oracle?
>
> Or, if possible, would it be best to have the shutdown script first
> try a shutdown immediate and then, if that doesn't work, try a
> shutdown abort?
>
>
> I think one disadvantage of a shutdown immediate is that it could
> cause corruption since it performs additional processing on the system
> that has bad hardware. So best to do a shutdown abort to get off the
> bad hardware system right away before any more damage is done. Does
> that make sense?
>
> And a disadvantage of a shutdown abort is that IF the active redo log
> is also corrupted (though this is unlikely), then the "instance
> recovery" on the companion system will fail. Is that right? (and how
> to recover from that? rollforward from a checkpoint???)
>
> Anyone with ideas or experience in this hardware-based system
> failover/Oracle shutdown scenario?
>
> Thanks
Received on Mon Jun 16 2003 - 07:31:20 CDT