RE: Filesystem for Archive logs in RAC
- From: "Mark W. Farnham" <mwf@xxxxxxxx>
- To: <tim@xxxxxxxxx>, <anuragdba@xxxxxxxxx>, "'ORACLE-L'" <oracle-l@xxxxxxxxxxxxx>
- Date: Thu, 1 Sep 2011 04:53:38 -0400
I tend to agree with Tim on the more important question, but I suppose there
are legitimate reasons for not including some things in your ASM managed
media farm.
If you have some reason (say a bunch of slower disks you already own and
don't want to mix in your ASM farm, or a filer that isn't up to spec for
being part of ASM), then it is possible to mount different directories r/w
on exactly one node versus read only on other nodes so that you can mount
the archived redo logs for a failed machine's instance to another machine
for recovery. If you engage in this topology, and your alert logs and
diagnostics are also not on clustered storage, you probably want to arrange
for them to be read only mountable also so you can look for a clue about
what went wrong on the node that is now down.
You may need to un-mount and re-mount to see the most recent files created.
I can't remember whether Solaris handles that transparently (it had to do
with re-reading the updated inode structure where this was a problem, but
I'm too lazy to look up the details). This is an antique idea from before
ASM existed, but you can do it to leverage non-clusterable storage or
clusterable storage without cluster license and/or overhead. If your number
of machines is greater than 2, then you'll probably want some protocol for
deciding which machine will mount and recover any particular failed node. It
is probably useful in some roll your own standby recovery systems where you
might want the simplest possible access to files by system utilities.
mwf
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Tim Gorman
Sent: Wednesday, August 31, 2011 5:09 PM
To: anuragdba@xxxxxxxxx; ORACLE-L
Subject: Re: Filesystem for Archive logs in RAC
Anurag,
You can certainly write archived redo log files out to any old
locally-mounted file-system as you like while the database is up and
running. The problem will occur when you need to use those files in a
recovery and the server on which they were mounted is not available, thus
neither are the archived redo log files residing there. Upshot: not a good
idea. So, to answer your question directly, you'd need either an
NFS-mounted file-system (such as NetApp) or a clustered file-system.
Now the more important question: why do you choose to put the archived
redo log files in a filesystem? What's wrong with a "flash recovery area"
disk group in ASM?
Hope this helps...
-Tim
-----Original Message-----
From: Anurag Verma [mailto:anuragdba@xxxxxxxxx]
Sent: Wednesday, August 31, 2011 02:47 PM
To: 'ORACLE-L'
Subject: Filesystem for Archive logs in RAC
I am in the process of configuring 11gR2 RAC on Solaris.
We are going to use ASM for shared storage for database files.
For Archive log files, we are going to put them in filesystem.
So in order to put the files in the filesystem, can you mount the same
filesystem (from SAN) on both RAC nodes same time,
so that RAC instances can write the archive logs in to the same directory?
How you guys have done at your setup?
Did you with filesystem for archive logs ? If yes, how did you do that?
I remember when I worked at one of my previous company, they were using
NetApp filers and we were using
filesystem for archive logs. Dont remember exactly how the storage admin did
that.
--
Anurag Verma,
Database Administrator
ERCOT(Electric Reliability Council of Texas),
Texas 76574
--
//www.freelists.org/webpage/oracle-l
Other related posts:
