Re: Shutdown & Datafiles - Sanity check

Anthony Hogan <ahogan-nospam-_at_iespana.es> wrote in message news:<3F32045E.FC8C8A88_at_iespana.es>...
> I have a small production database 8.1.7 running on W2K server. I run a
> nightly script to shut the database down and os copy the datafiles. The
> alert file shows no errors for the shutdown or subsequent startup.
> The curious thing is that the os timestamp on the temp datafile is not
> in sync with the rest of the datafiles. DBV shows no errors,
> v$datafile_header.checkpoint_time is correct and a restore to another
> server completes successfully. So... what's going on?
> My understanding is that on shutdown the datafiles should all have the
> same timestamp. Is this the case or am I being unduly concerned?
>
> TIA
> Anthony Hogan

If you are using your tempfile for sorting (which is a normal thing), it will be written to when a sort is too big to fit into memory. So if no sort ever gets that big, it will have the origination timestamp.  If you have the memory to make your sort area in the SGA bigger, that could help performance on those tings that require sorting (perhaps imports, index creation as well as batch type operations and reports).  See v$sysstat sort info, and v$sort_mumblesomething if you are really curious.

As others have pointed out, it isn't really a datafile.

Of course, I don't have much experience on Windows so I might be totally wrong. It certainly seems to work that way on unix. The sort segments are supposed to be cleaned up on startup by SMON, but I've seen that work very strange (as in "not" or "halfway") sometimes.

jg

--
@home.com is bogus.
"The question is: What is happening to the total IT budget?  I think
it's going to shrink. We're down to three--IBM, Microsoft and Sun. The
rest is collateral damage."  - Scott McNealy