Re: Shutdown Abort - Not a Good Idea, Right?

Here is what happens:

    The checkpoint counter in the data file header is incremented each time Oracle performs a checkpoint on that data file. The controlfile keeps a checkpoint counter for every data file as well. Also every data file header contains an SCN called the start SCN. The controlfile has an SCN for every data file called the stop SCN. As with the checkpoint counter, the start SCN for every data file is incremented each time a checkpoint is done. During normal database operation the stop SCN in the controlfile is set to infinity.

    If the database is shutdown normal or immediate, the checkpoint issued sets the corresponding stop SCNs in the controlfile to the start SCN currently in each data file. Next time you open the database Oracle makes two comparisons. First it sees if the checkpoint counter in the data file header matches the corresponding checkpoint counter in the controlfile. If yes, it does the second check. This one sees if the start SCN in the data file header is the same as the corresponding stop SCN in the controlfile. If yes, then both tests have been passed and no recovery is required. Every data file is looked at, then the database is opened. As the database opens the stop SCNs in the controlfile are reset to infinity.

    But suppose the database crashes or you issue a shutdown abort. In this case a checkpoint is not issued and the stop SCN in the control file remains at infinity. During the next startup Oracle makes two comparisons. If the checkpoint counters in the data file header and control file match (i.e. you didn’t replace the data files with a backup copy) you pass the first test and move on to the second. In the second test the start SCN will have some value but the stop SCN will be set to infinity, so they won’t match. Oracle will decide that crash recovery needs to be performed.

    There have been pros and cons in the messages above. Oracle recommends that you not do a shutdown abort as part of a routine backup. I would pay attention to their recommendation.

Van

M. Bhatti wrote in message <371DEA8B.8EB8C2BD_at_mci.com>...
>I just recently noticed that, before doing cold backups, our db
>instances (Oracle 7.3.2/8.0.5, Sun Solaris 2.5.1) are shutdown aborted
>than brought up and shutdown normal, so that the cold backup can be
>started immediately after the shutdown normal.
>
>Question is, I thought shutdown abort should not be used except in
>extreme conditions ie when a db instance
>cannot be shutdown normal/immediate.
>
>So, I assume that:
>This is bad idea right (shutdown abort)?
>What can happen to the db?
>
>Thanks for any input.
>
>mkb
>
>
Received on Thu Apr 22 1999 - 19:54:07 CDT