Re: Shutdown Immediate Command

robertgfreeman_at_my-deja.com wrote:
> I suppose, in this case, if one supposes that the problem with the
 redo
> log is one of corruption that is occuring after the commit completes,
> and that Oracle is not picking up on the corruption, that if you
> shutdown abort and then startup the database, that instance recovery
> could fail because of said corruption. You would end up with an
> inconsistent database and would have to use one of the fun _*
> parameters to force it open!

Great point about corrupted redologs, but you'd have to have a seriously malfunctioning OS for that to happen as you describe. If LGWR encounters an error performing a write() call, it will crash the instance without returning a successful commit to those transactions that are part of the log buffer flush taking place. If something is going around corrupting redo after LGWR writes it, well, you have a bigger problem than just crashes. Oracle couldn't perform automatic cache repairs, and all your backups would be useless.

> I've seen problems with SHUTDOWN ABORTS in 7, but then only with
> databases in NOARCHIVELOG mode. I know that Oracle in Oracle8 will not
> allow you to overwrite a online redo log in which the dirty buffers
> related to the RBA of the log, even in ARCHIVELOG mode.

Of course - this is a basic assumption at least from v.7 forward. I don't know about before that. You can't write to a redolog unless a checkpoint has completed since LGWR switched out of that log.

So, what were the problems you has with 7 and shutdown abort?

> I've never seen a problem with databases in ARCHIVELOG mode in terms
 of
> SHUTDOWN ABORT, as long as archived redo logs and online redo logs are
> available.

How would having archivelog either enabled or disabled have any effect on crash recovery and shutdown abort? Crash recovery only uses online redologs. As you just pointed out, the basic premise of online logfiles is that all the datafiles are checkpointed before Oracle reuses them. Where would archived redologs come into play for a crash recovery?

--
Jeremiah


Sent via Deja.com
http://www.deja.com/