Fwd: Re: What do others do for database file mounts

  • From: Wayne Smith <wts@xxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Wed, 18 Feb 2015 12:01:09 -0500

I meant this for the group...
---------- Forwarded message ----------
From: "Wayne Smith" <wts@xxxxxxxxx>
Date: Feb 18, 2015 12:00 PM
Subject: Re: What do others do for database file mounts
To: <rjoralist3@xxxxxxxxxxxxxxxxxxxxx>
Cc:

You'll want to have an understanding of the performance of various disk
spaces your storage people can allocate.  Some of your NetApp may have many
smaller disk and some fewer very large disks.  I want my  redo logs in the
former,  generally.   Some space may have differing performance due to
cache. Some will have differing performance due to your sharing space,
under the covers,  with other activity.  This will vary over time!
Performance will not be constant as days and months proceed.   Also,  the
other applications will have bursts of activity and calm.  Most of all,
perhaps,  if NetApp dedup is turned on,  you may experience significant
performance degradation during the scheduled dedup times.

Your space may actually move,  under the covers, as your storage admin
manages everyone's space (assuming you have the new "cluster" capability
enabled).

We're moving most of our space from lun to NFS (dNFS),  for manageability.
Thin provisioning,  auto grow,  and online growth are wonderful.

One thing to be aware of is how and where space for snapshots is
accomplished.  Frequent automatic snapshots on an active volume can take a
lot of space,  and with  NFS,  the space may be configured to come from the
allocation your op/sys sees.   It was really weird to see my redo space
fill,  just because the database became more active (snapshots were on by
mistake).

In general,  the only non RMAN backup of my database spaces is
/oracle/diag.

To the more direct question of mounts,  I like nothing interfering with
redo and redo on 2 separate everything and very fast write. I like archive
redo about the opposite..  cheap,  performance not a concern,  but with 2
separate destinations, if possible.  Control files and redo don't mix in my
book  ;-)

Cheers, Wayne
On Feb 13, 2015 5:06 PM, "Rich Jesse" <rjoralist3@xxxxxxxxxxxxxxxxxxxxx>
wrote:

> Andy posts:
>
> > We are moving to Fibre Channel connections with all of our storage on a
> Net
> > App storage unit.  One of the questions we got from our system admins
> was if
> > we could limit the number of mounts that they provide us for out
> datafiles.
> > With the newer disk technology that we all use, is it still important to
> > segregate mounts for redo, CFs, datafiles, Archive logs, etc?  I know
> Oracle
> > stands firm that best practices are to segregate these files into
> separate
> > mounts.  I’m wondering what others do and if there is a compelling
> reason
> > to continue this practice?
>
> The main reason I kept the split on a new AIX v7.1 server is because the
> old
> AIX 5.3 server had experienced JFS2 filesystem corruption.  Luckily this
> was
> easily correctable with a fsck, but it scared me enough to still keep
> everybody separated, including mountpoints for each of the three
> controlfiles.  This is on an LPAR, so all storage is on the SAN (XIV).
>
> Also, in a disk space emergency, I don't want to worry about trying to
> figure out what could be filling a mountpoint that's shared by redos,
> archived redos, RMAN backups, MP4s of kittens, etc., or the ramifications
> to
> all of the above should the MP fill completely.  One bucket per item makes
> targeting potential problems and their respective solutions easier.
>
> For this solo DBA, at least.  YMMV.
>
> My $.02,
> Rich
>
>
>
> --
> //www.freelists.org/webpage/oracle-l
>
>
>

Other related posts:

  • » Fwd: Re: What do others do for database file mounts - Wayne Smith