Re: The weather in Scotland. (Oracle is robust).

"George Barbour" <gbarbour_at_csc.com> wrote in message news:<3d4914e7$1_at_pull.gecm.com>...
> >Morning George,
> >fine weather we are having just now eh ?
> >Cheers,
> >Norman.
> Norman made this comment in the bookies thread, I thought I'd take it out of
> context and shove it here.
> What Norman is referring to, is the "American Style" Thunder and Lightning
> we have had in Scotland (and Englandshire) for the last few days; the stuff
> we see on documentaries about the wonders of weather in America. Never seen
> anything like it!
> Anyway, it caused our server systems to crash big time, mostly because of
> severe power outages.
> But what I want to say here, is that the Oracle8i systems performed
> superbly, even though they had to put up with complete black outs, they all
> restarted, recovered automatically (thread recovery, crash recovery) and
> continued as though nothing had happened!
> Thank you Oracle, superb product. I was worried in case I actually had use
> the backup system I had put in place.
> BTW the Access folks are in real trouble.
>
> George Barbour.
> PS Yes I have tested the backup system using different recovery scenarios
> etc.

Yes, I have had similar fortune. I hate running recoveries. I was running forward recovery once when Oracle reported corruption in the archived redo log and refused to recovery any further. I believe that at least in 9i Oracle will now allow you skip the bad block and continue, but back then we were dead. Fortunately all data tablespaces, rbs, and system were current so my need recovery datafiles were part of an index tablespace. I had to generate hundreds of indexes from the dictionary, drop the datafiles, start Oracle in recovery mode, and recover without the index tablespaces. Recreate the missing tablespaces and rebuild the indexes. I was fortunate a data tablespace was not involved as setting the database backwards in time is not an acceptable option.

It is always better when the db just recoveries itself and you do not have to test your recovery because you never know how old that tape System Administration is using for your backup happens to be. Or how about this one, the System Admin did not check their backup logs and the backup was filling the tape, rewinding, then continuing on overlaying part of the backup.

Sometime it is wise to ask if other people involved in the backup chain have physically read any of this week's backup logs just to be sure everything is OK. You might take a little static for making the request but if your shop relies on manual procedures or experience regular problems with the tape management system I find it to be worth the effort to pester people to do their jobs or point you to the logs so you can look. More than once I have found problems that could have cost us dearly.

• Mark D Powell --