Recovery problem with oracle 6

From: DSS Unix Support <support_at_jfdi.demon.co.uk>
Date: Sun, 14 Aug 1994 12:25:29 +0000
Message-ID: <776867129snz_at_jfdi.demon.co.uk>


I'd appreciate comments and/or advice on a problem that I hit this weekend on an HP 9000 running Oracle 6.

We had to patch a bad disc, so took a full backup of all the Oracle filesystems (binaries and data) onto 2 DAT tapes. Having an hour spare, we then copied the filesystem with the bad block (/apps) onto a separate tape in an effort to minimise the recovery time.

Patched the disc & rebooted the system bringing Oracle and it's application up automatically (a mistake, I know). Closed Oracle, and brought back /apps from the individual DAT (the third tape, produced in the spare hour).

Tried to start Oracle to find that one of the rollback files failed the validation check.

Brought back the full system from the 2 DAT tapes, started Oracle and got exactly the same error.

The root cause of the problem (as I understand it) is the timestamps held inside the various Oracle data files not being in step, and this was caused by a combination of starting Oracle after patching the disc, and using HPUX's frecover without the -o option (which would have forced recovery of files older than those currently on disc). The second, full recovery attempt (again, no -o!) replicated the problem throughout all the Oracle filesystems.

Given that my assessment of the cause is correct (feel free to correct it if you disagree), is there any way of forcing Oracle to override the timestamp checks in it's files? I realise that in normal circumstances this would be dangerous, but is it possible? Given the situation that I had, was there an alternative recovery path I could have tried?

Thanks,

Paul Wilson

-- 
CWS Unix Support
Received on Sun Aug 14 1994 - 14:25:29 CEST

Original text of this message