Questions about catalog archiving / converting (nb5.1)

[Nathan_Kippen]

Just wondering if anyone could share their experience with either converting their catalog to binary format or archiving.  I have a catalog now that now takes about 3.5 hours to backup.  I'm using DLT600 tapes, but I'm afraid some day I'll soon run out of space.  (Of course there's the option of using multiple tapes, but that's not the point.)

Also .. right now we only use a master/media server (1 server total).  What are the advantages of having several media servers if my master server can "handle" everything up to this point.  We'll be getting an Itanium soon to replace our current master, but anyway, if the master server can handle it is there really a need for media servers?

Thanks...

(p.s. if there is any white paper or something that could shed light on subject that would be good too)

[Casey_King]

I am not sure if you have this pdf, but do a search form Veritas Netbackup Planning and Performance Tuning Guide.  I believe the current version is Release 6.0.

3.5 hour catalog backup seems quite long.

My catalog backup runs for 2 hours and I have 8 media servers, but sadly only on LTO1 drives (will be ugrading drives soon to LTO3 so I get get 2GB to the drive, the rest of my backup network is 2GB).

[Omar_Villa]

Hi take a look on this: http://support.veritas.com/docs/280022  Im sure it will give you the guide you need.

Hope this helps.

Regards

[Chia_Tan_Beng]

---

Nathan Kippen wrote:

Also .. right now we only use a master/media server (1 server total).  What are the advantages of having several media servers if my master server can "handle" everything up to this point.  We'll be getting an Itanium soon to replace our current master, but anyway, if the master server can handle it is there really a need for media servers?

You may want to consider that when you'll have new media server(s) joining this backup domain and your backup domain is very busy. I "moved" my master server out of the master/media server configuration for one main reason, to isolate any media server related issues that requires reboot. In my case, the backup operation runs 24hrs and some sticky media server problem (drive/library/SAN related) required a server roboot. With the master server component being separated, the reboot would not affact the entire backup domain.    

[sdo]

Hi Nathan,

Env: NBU ES v5.1 MP6, master MSCS cluster 2 x Win2000 Adv Srvr SP4, 16 x LTO1 SSO, 3 x Win media plus 2 x Solaris 8 media.

Our catalog backup of 160GB data on an NTFS SAN volume is compressed by NetBackup to 100 GB on disk.  Our catalog backup was creeping up to between 5:45 and 6:00 hrs regularly.  We implemented RaxCo PerfectDisk 8, and after two days of running defrags during quite periods (after catalog and before main schedules) we got our catalog backup back down to between 3:30 and 3:45.

Of course I would still expect faster catalog backups, but my bare bones bpbkar32 disk read speed tests have confirmed that using "compression" on a Windows 2000 based NetBackup Catalog just slows disk reads in my environment to a sustained 10MB/s.  I can't make my catalog backups quicker unless we go to larger SAN attached volume for the catalog so that I can then uncompress the catalog.

In summary: It certainly pays to defrag NetBackup catalog volumes if you are using NTFS on Windows master servers.

Regards,

Dave.

P.S. We have found RaxCo PerfectDisk v8 to be stable and safe to run during normal quietish operations (i.e. during restores, the odd ad-hoc backup, and during the usual catalog sweeps by the image expiry and catalog compress background services/daemons/jobs).  We haven't tried to run a defrag during actual production backup runs - but we haven't had to stop anything either - and no errors in either Raxco logs, NBU logs or Event logs.  We've been running this for six months now without any problems.

[Nathan_Kippen]

Thank you everyone for your responses.  It will all be taken into consideration as I move towards 6.0