Re: LGWR and DBWR

On 25 feb, 19:03, joel garry <joel-ga..._at_home.com> wrote:
> On Feb 25, 4:33 am, Helma <helma.vi..._at_hotmail.com> wrote:
>
>
>
> > On 25 feb, 02:52, "Dereck L. Dietz" <diet..._at_ameritech.net> wrote:
>
> > > Oracle 10.2.0.3.0
> > > Windows 2003 Server
>
> > > Is it possible for improperly sized redo log files to be overwhelmed to the
> > > point where it will cause a process which is performing a lot of writes to
> > > multiple tables to stop processing without any error or reason in a log
> > > file?
>
> > No.
>
> > With overwhelmed redo logfiles the error in the alert logfile would
> > be:
>
> > Thread 1 cannot allocate new log, sequence 50776
> > Checkpoint not complete
>
> Perhaps I'm misreading the OP, but isn't this the effect one sees when
> all redo's are full and Oracle can't switch to the next one?
> _Everything_ stops until it can switch... if the OP means just one
> process of many, that would be different.
>
> jg
> --
> @home.com is bogus.http://www.csmonitor.com/2008/0225/p01s02-usgn.html

> Perhaps I'm misreading the OP, but isn't this the effect one sees when
> all redo's are full and Oracle can't switch to the next one?

Yes, that is my interpretation of his question. The redologfiles will fill up because of to slow IO to the datafiles. His question was if such an event could happen WITHOUT ANY ERROR OR REASON IN A LOGFILE.
This is not true, the above message will be send to the alertlogfile. He seems to have a problem with slow IO and wonders if inapropriately sized redo logs could be a reason. The answer should be : not without the 'cannot allocate new log' message.

If there is a different interpretation of his question, let us hear it.

H. Received on Tue Feb 26 2008 - 05:59:06 CST