|
|
 |
 |
 |
 |
Backup to NFS
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) ....... We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) remove NSPAM to email
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) > ....... > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > remove NSPAM to email
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
-----------------------------------------------Reply-----------------------------------------------
On May 8, 10:27 am, sybrandb <sybra @gmail.com> wrote:
> 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) > > ....... > > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > > remove NSPAM to email > 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
why? -----------------------------------------------Reply-----------------------------------------------
On May 8, 11:08 am, Cristian Cudizio <cristian.cudi @yahoo.it> wrote:
> On May 8, 10:27 am, sybrandb <sybra @gmail.com> wrote: > > 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) > > > ....... > > > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > > > remove NSPAM to email > > 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 > why?
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
-----------------------------------------------Reply-----------------------------------------------
On May 8, 11:11 am, Cristian Cudizio <cristian.cudi @yahoo.it> wrote:
> On May 8, 11:08 am, Cristian Cudizio <cristian.cudi @yahoo.it> > wrote: > > On May 8, 10:27 am, sybrandb <sybra@gmail.com> wrote: > > > 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) > > > > ....... > > > > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > > > > remove NSPAM to email > > > 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 > > why? > I've never tried, but i'm just curious to undertand what problems may > arise. > thanks. > Cristian Cudizio > http://oracledb.wordpress.comhttp://cristiancudizio.wordpress.com- Hide 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
-----------------------------------------------Reply-----------------------------------------------
On May 8, 11:21 am, sybrandb <sybra @gmail.com> wrote:
> On May 8, 11:11 am, Cristian Cudizio <cristian.cudi @yahoo.it> > wrote: > > On May 8, 11:08 am, Cristian Cudizio <cristian.cudi@yahoo.it> > > wrote: > > > On May 8, 10:27 am, sybrandb <sybra@gmail.com> wrote: > > > > 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) > > > > > ....... > > > > > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > > > > > remove NSPAM to email > > > > 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 > > > why? > > I've never tried, but i'm just curious to undertand what problems may > > arise. > > thanks. > > Cristian Cudizio > >http://oracledb.wordpress.comhttp://cristiancudizio.wordpress.com-Hide 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 http://oracledb.wordpress.com http://cristiancudizio.wordpress.com
-----------------------------------------------Reply-----------------------------------------------
On May 8, 2:27 am, sybrandb <sybra @gmail.com> wrote:
> 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) > > ....... > > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > > remove NSPAM to email > 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
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.
-----------------------------------------------Reply-----------------------------------------------
On May 8, 12:21 pm, Cristian Cudizio <cristian.cudi @yahoo.it> wrote:
> On May 8, 11:21 am, sybrandb <sybra @gmail.com> wrote: > > On May 8, 11:11 am, Cristian Cudizio <cristian.cudi@yahoo.it> > > wrote: > > > On May 8, 11:08 am, Cristian Cudizio <cristian.cudi@yahoo.it> > > > wrote: > > > > On May 8, 10:27 am, sybrandb <sybra@gmail.com> wrote: > > > > > 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) > > > > > > ....... > > > > > > We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) > > > > > > remove NSPAM to email > > > > > 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 > > > > why? > > > I've never tried, but i'm just curious to undertand what problems may > > > arise. > > > thanks. > > > Cristian Cudizio > > >http://oracledb.wordpress.comhttp://cristiancudizio.wordpress.com-Hid... 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 > http://oracledb.wordpress.comhttp://cristiancudizio.wordpress.com- Hide 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. -- Sybrand Bakker Senior Oracle DBA
-----------------------------------------------Reply-----------------------------------------------
On Tue, 08 May 2007 01:27:26 -0700, 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.
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. -- http://www.mladen-gogala.com
-----------------------------------------------Reply-----------------------------------------------
Fuzzy wrote: > On May 8, 2:27 am, sybrandb <sybra @gmail.com> wrote: >> 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) >>> ....... >>> We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) >>> remove NSPAM to email >> 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 > 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.
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 damor @x.washington.edu (replace x with u to respond) Puget Sound Oracle Users Group www.psoug.org -----------------------------------------------Reply-----------------------------------------------
Mladen Gogala wrote: > On Tue, 08 May 2007 01:27:26 -0700, 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. > 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.
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. -- Daniel A. Morgan University of Washington damor@x.washington.edu (replace x with u to respond) Puget Sound Oracle Users Group www.psoug.org
-----------------------------------------------Reply-----------------------------------------------
On Tue, 08 May 2007 08:28:43 -0700, DA Morgan <damor @psoug.org> 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. -- Sybrand Bakker Senior Oracle DBA
-----------------------------------------------Reply-----------------------------------------------
On Tue, 08 May 2007 08:30:36 -0700, DA Morgan wrote: > 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 cant 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. -- http://www.mladen-gogala.com
-----------------------------------------------Reply-----------------------------------------------
sybra @hccnet.nl wrote: > On Tue, 08 May 2007 08:28:43 -0700, DA Morgan <damor @psoug.org> > 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.
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. -- Daniel A. Morgan University of Washington damor@x.washington.edu (replace x with u to respond) Puget Sound Oracle Users Group www.psoug.org
-----------------------------------------------Reply-----------------------------------------------
On 8 May 2007 01:27:26 -0700, sybrandb <sybra @gmail.com> wrote: >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 ....... We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc) remove NSPAM to email
-----------------------------------------------Reply-----------------------------------------------
On Tue, 08 May 2007 20:52:13 +0200, sybrandb wrote: > 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. -- http://www.mladen-gogala.com
|
 |
 |
 |
 |
|