Re: NFS mounted stale redo log

From: joel garry <>
Date: Tue, 12 Feb 2008 12:55:43 -0800 (PST)
Message-ID: <>

On Feb 12, 10:13 am, rjlolli <> wrote:
> We have come across a strange situation concerning redo logs that are
> NFS mounted.  We have an Oracle database with 4 redo log groups with 3
> members in each group.  Of the 3 members, one of those member is NFS
> mounted remotely.  From time to time, Oracle will mark the NFS mounted
> redo log member stale for a split second and then it's status will go
> back to normal.  I had always thought once Oracle marks a redo log
> stale it remains stale until Oracle cycles back around through log
> switches.  Why is Oracle doing this?  Is there a parameter that will
> assure us the member remains stale?  My theory is that the NFS mounted
> redo log is marked stale for such a short period of time that Oracle
> is re-synching it with the remaining 2 local members and moving on.
> Please help if you have any ideas.  Thanks.
> BTW.  This is on Oracle 9.2

Four questions:

Are you using a supported NFS, properly configured?

Are you seeing checkpoint incomplete messages?

Do you have very small or very large redo logs and log buffers?

What are your top waits?

I'm speculating you are seeing the effects of a slow checkpoint, perhaps caused by too much writing to the logs (very small logs might mean too many writes by dbwr blocking the lgwr, very large might mean hitting an i/o bottleneck during archiving, something similar for log buffer waits).


-- is bogus.
40 years on, and Black Dog is now country music.
Received on Tue Feb 12 2008 - 14:55:43 CST

Original text of this message