Re: The delay in backup controlfile as trace?

On Oct 27, 9:36 am, "Ultra Kiasu" <ultraki..._at_hotmail.com> wrote:
> Using Oracle 9i, on HPUX IA64 11.23, Hitachi HDS Storage system.
>
> Sometimes, when my DBA is performing a online hotbackup, the backup
> controlfile trace file is missing. We did a manual trace, and
> discovered that the tracefile creation is delayed, even the system
> returned the message "Database altered"..
>
> Alter database backup controlfile to trace;
>
> Database altered
>
> Even sometimes, the current redo logs are not flushed to archive logs
> immediately when issue command "alter system switch logfile;"
>
> Anyone know how to minimise this delay?
>
> We are running a quick shadow image using the Hitachi HDS storage
> system, the snapshot backup take no longer than 2 mins to complete, but
> the creation of control trace files or flushing of archivelog take much
> longer than 5 mins to complete.
>
> Thus render my backup incomplete, or useless for recovery purpose, and
> inserting a hardcoded delay is not really productive.
>
> Any solution?

As you are using non-standard backup mechanism, and there is simply no such delay in Oracle, I suggest this is a problem with your config and *not* an Oracle problem.
Also: are you using ordinary ufs file systems? In that case probably the interaction with your filesystem cache is the culprit. In that case I would look into tuning the O/S and/or making sure you mount the filesystems with the correct options or switch to raw disk. You could of course also run a trace on event 10046 level 8, to find out what Oracle is waiting for.

-- 
Sybrand Bakker
Senior Oracle DBA

Sybrand