Backup to NFS
NetComrade
achieve. We'd like to be able to backup multiple db's at once, so
something in the range of 100Megs+/sec is what we are looking for
(obviously for cheap). Storage size is a few TB.
(need to buy a new one)
.......
We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc)
remove NSPAM to email
sybrandb
wrote:
Did you ever try to restore from such a NFS backup?
I did. It wasn't funny. I never got the database back.
--
Sybrand Bakker
Senior Oracle DBA
Cristian Cudizio
why?
Cristian Cudizio
wrote:
I've never tried, but i'm just curious to undertand what problems may
arise.
thanks.
Cristian Cudizio
http://oracledb.wordpress.com
http://cristiancudizio.wordpress.com
sybrandb
>
> - Show quoted text -
The savesets were incomplete, causing several datafiles NOT to
restore. IIRC, one of them was the SYSTEM tablespace.
Note: Absolutely no sign of any problem during backup!
Cristian Cudizio
>
> > - Show quoted text -
>
> The savesets were incomplete, causing several datafiles NOT to
> restore. IIRC, one of them was the SYSTEM tablespace.
> Note: Absolutely no sign of any problem during backup!
>
> --
> Sybrand Bakker
> Senior Oracle DBA
But do you think it was caused by NFS? Similar problems arised to me
with backup on
NTFS on windows machines (corrupted backups).
Thanks,
Fuzzy
Which NFS were you using?
If the traditional one based on UDP, I concur that you could/should
expect challenges. However, I'd expect most propblems to disappear
when using the new NFS based on TCP.
I've done some of this kind of backup to NetApp Filer using their NFS
and had no challenges. Then again, their NFS is also certified for
Oacle RAC.
--
Hans Forbrich (mailto: Fuzzy.GreyBeard_at_gmail.com)
*** Feel free to correct me when I'm wrong!
*** Top posting [replies] guarantees I won't respond.
sybrandb
>
> > > - Show quoted text -
>
> > The savesets were incomplete, causing several datafiles NOT to
> > restore. IIRC, one of them was the SYSTEM tablespace.
> > Note: Absolutely no sign of any problem during backup!
>
> > --
> > Sybrand Bakker
> > Senior Oracle DBA
>
> But do you think it was caused by NFS? Similar problems arised to me
> with backup on
> NTFS on windows machines (corrupted backups).
>
> Thanks,
>
> Cristian Cudizio
>
>
> - Show quoted text -
What I did observe was a backup to local disk and to tape restored
flawlessly, a backup to NFS didn't.
I never had any problems restoring a backup from NTFS drives (yes, I
did try), although obviously those were not networkdrives.
Mladen Gogala
> Did you ever try to restore from such a NFS backup? I did. It wasn't
> funny. I never got the database back.
I must say I've had no such problems. The only problem I had was speed.
Local backup/restore operations were much faster then the NFS ones.
NetApp was still much better then a direct backup to tape using Veritas.
As long as the remote share is properly mounted, I was able to
backup/restore. The meaning of "properly mounted" is described in the
Metalink note 359515.1.
DA Morgan
Same experience here. If NetApp's NFS, AFAIK the only one certified by
Oracle, I've never experienced a problem.
--
Daniel A. Morgan
University of Washington
damo...@x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org
DA Morgan
An important consideration I should have mentioned before. RMAN and a
temporary NFS mount such as:
# mount -t nfs 192.168.20.160:/vol/stage /mnt
will not work: Mount using FSTAB.
sybr...@hccnet.nl
wrote:
>Same experience here. If NetApp's NFS, AFAIK the only one certified by
>Oracle, I've never experienced a problem.
I'm not talking about using NetApp NFS. I'm talking about NFS supplied
by the O/S. Does that mean O/S supplied NFS (even v3 or higher) is not
supported? In several situation people just force me to use NFS
because of insufficient diskspace.
Mladen Gogala
> An important consideration I should have mentioned before. RMAN and a
> temporary NFS mount such as:
>
> # mount -t nfs 192.168.20.160:/vol/stage /mnt
>
> will not work: Mount using FSTAB.
Well, I believe that options are more important then the technique itself.
Configuration files like /etc/fstab are also parsed and, eventually,
handed to a program that calls "mount" system service - see mount(2)
or even the CLI version of the mount. During the startup Unix systems
do a horrible patchwork of grepping through /etc/fstab, /etc/hosts and
other configuration files. Newer versions of Linux even have Mother's
little helper like this:
[root@medo ~]# which fstab-decode
/sbin/fstab-decode
[root@medo ~]#
[root@medo ~]# file /sbin/fstab-decode
/sbin/fstab-decode: ELF 32-bit LSB executable, Intel 80386, version 1
(SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.9,
stripped
[root@medo ~]#
[root@medo ~]# man fstab-decode
fstab-decode(8) fstab-
decode(8)
NAME
fstab-decode - run a command with fstab-encoded arguments
SYNOPSIS
fstab-decode COMMAND [ARGUMENT]...
DESCRIPTION
fstab-decode decodes escapes in the specified ARGUMENTs and uses
them
to run COMMAND. The argument escaping uses the same rules as
path
escaping in /etc/fstab, /etc/mtab and /proc/mtab.
EXIT STATUS
fstab-decode exits with status 127 if COMMAND can’t be run.
Otherwise
it exits with the status returned by COMMAND.
EXAMPLES
fstab-decode umount $(awk ’$3 == vfat { print $2 }’ /etc/fstab)
This program is clearly intended for being used within the sysinit script.
My only gripe is that nobody thought of using Perl for that purpose. The
only thing missing in /etc/init.d/rc.sysyinit is Perl.
In other words, the effect of writing a monstrous command line like:
mount -t nfs -o
rw,bg,hard,nointr,rsize=32768,wsize=32768,tcp,vers=3,timeo=600 lap:/
export/fs /mnt
and putting it in /etc/fstab will be exactly the same, as far as oracle
is concerned.
DA Morgan
When you NFS mount it is always with the operating system.
But there is, it appears, a difference between:
# mount -t nfs 192.168.20.215:/vol/stage /mnt
and
ntap270a:/vol/alpha /u01 nfs
rw,bg,intr,hard,rsize=32768,wsize=32768,noac,nolock,tcp,vers=3 0 0
Both being to a NetApp filer head.
One is accepted by RMAN, RAC, etc. the other is not. And from
compatibility matrices it appears that Oracle distinguishes
NetApp from others. Again I don't know the details as to why.
NetComrade
>On May 8, 9:57 am, NetComrade <netcomradeNS...@bookexchange.net>
>wrote:
>> If you backup to NFS, which device you use, and what throughput do you
>> achieve. We'd like to be able to backup multiple db's at once, so
>> something in the range of 100Megs+/sec is what we are looking for
>> (obviously for cheap). Storage size is a few TB.
>>
>> (need to buy a new one)
>>
>
>Did you ever try to restore from such a NFS backup?
>I did. It wasn't funny. I never got the database back.
The reasons are usually economical. It's much cheaper to have a big
NFS sitting somewhere, then have local disk, which usually means
fibre, to be backing things up. (we backup to disk then to tape)
Yes we have restored many times. Yes, we did have a problem once,
which is why we now 'validate' those backups. And I agree with other
posters that you can have a corruption on a local level, just as you
can over NFS, I will also somewhat agree that there is a higher
probability with NFS. Provided you have a speedy enough NFS, and a
dedicated network, it should not be a problem though.
-a
Mladen Gogala
> I'm not talking about using NetApp NFS. I'm talking about NFS supplied
> by the O/S. Does that mean O/S supplied NFS (even v3 or higher) is not
> supported? In several situation people just force me to use NFS because
> of insufficient diskspace.
If you use the options specified in the document I gave you, you should
be fine. Of course, you should test.