Re: Recovery with logs, then incremental, then more logs?

From: "Binley Lim" <Binley.Lim@xxxxxxxxxx>
To: <Brandon.Allen@xxxxxxxxxxx>, "Riyaj Shamsudeen" <rshamsud@xxxxxxxxxxxx>
Date: Wed, 19 Apr 2006 14:10:36 +1200

Don't know about EMC specifics, but Netapp has a snapshot feature where you
are using up additional disk-space each time you do a DML. The original
block is "frozen" (contains the snapshot point-in-time), and the copy of the
block with the modified data is written out. So now you have 2 blocks
instead of one. As you make more changes, you will take up more space.

Fact that you got the no-space-left error when writing to the online redo
log sounds suspiciously like the same sort of problem, ie a redo write is
actually requiring more space. If this is a snapshot issue, keeping to 10%
free will of course avoid this problem. Better still, switch off
snapshotting.

Also it appears you have put the redologs on the same volume as the
archivelogs. I would prefer to keep them separate, like internal/DAS disks,
for these and other reasons vendors don't tell you about. And for
performance reasons, also look at increasing rsize/wsize - and EMC should
have recommendations on that.

----- Original Message -----
From: "Allen, Brandon" <Brandon.Allen@xxxxxxxxxxx>
To: "Riyaj Shamsudeen" <rshamsud@xxxxxxxxxxxx>
Cc: <oracle-l@xxxxxxxxxxxxx>
Sent: Wednesday, April 19, 2006 10:25 AM
Subject: RE: Recovery with logs, then incremental, then more logs?


> Yep, it's an EMC.  My Unix admin just told me earlier today that EMC told
him that they recommend keeping the filesystems below 90% full to avoid any
problems.  I had never heard anything like that before - anyone else?  This
is a 500GB filesystem, so to keep it below 90% we're talking about wasting
50GB.  Not the end of the world, I know, but certainly not ideal and I can't
believe that this would actually be a requirement for stability.
>
>
> -----Original Message-----
> From: Riyaj Shamsudeen [mailto:rshamsud@xxxxxxxxxxxx]
>
>
> I don't know much about NFS3 protocol, but is your vendor / file system
> in this list ?
>
>
http://www.oracle.com/technology/deploy/availability/htdocs/vendors_nfs.html


--
//www.freelists.org/webpage/oracle-l

References:
- RE: Recovery with logs, then incremental, then more logs?
  - From: Allen, Brandon

Re: Recovery with logs, then incremental, then more logs?

Other related posts: