03-11-2014 12:57 PM
Hi there,
Our catalog backups were running fine, and now I find that nebackup is using two tapes for a catalgoue backup.
We are running 7.6.0.1.
The first task writes 500 MB to one tape, than another task kicks off which writes 2 GB to another tape. We are using LTO4 media. Prior to last week, catalog backups were utilizing one volume.
Can anyone offer any insight into this behaviour?
Thanks
Seth
Solved! Go to Solution.
03-13-2014 03:33 AM
I dont think this is a 7.6 thing as it has been known to happen in other environments as per the link Andy Welurn supplied earlier which reflected exactly the same thing.
It generally comes down to tape drive load balancing which is built into NetBackup
NetBackup tries to balance the number of mounts each drives has so if you fire off two jobs, one after the other, the first will go to the first drive and the second will go to the second drive just to keep number of mounts on each drive the same.
What actaully happens is often down to pure luck .. if while the first part of the catalog backup is running the other drives get plenty of mounts then it will carry on using the same drive, if not it will swap.
It all seems very random but there is some sort of logic behind it somewhere and the only way i have seen to make sure it doesn't happen at some point is to use the command as per the linked thread to allow only one piece of partially full media - though you may actually see it be moved from the first drive and mounted in the second drive if NetBackup thinks that is best.
Hope this helps, let me know if:
vmpool -update -pn CatalogBackup -mpf 1
still works in 7.6
03-11-2014 03:24 PM
How about I just tell you how to stop it ...
Find the list of volume pools, the catalog backup wil have a 'special pool' callcatalog, or similar.
Right click this and select properties / change / options (apologies, can;t remember the exact menu name).
In there you can see the setting for 'Max partially full volumes" - set this to 1
It might have to finish the two tapes it is currently using, but after one of them becomes full, it will in the future only use one tape at a time.
As to why, if it happened after an upgrade, I can only image something has changed in the code if no other changes have been made.
03-12-2014 01:07 AM
Another option which *may* alleviate this - create a separate Storage Unit specifically for your Catalog backups & restrict the number of drives on that to one.
We have had this issue & it has come up on the forum many times in the past and either this or Martins suggestion has been employed to 'correct', altho' Marks solution in the following post indicates that there is (was) a round-a-bout way of applying the mpf level on a catalog volume pool (not sure if this is still the case)
https://www-secure.symantec.com/connect/forums/can-single-catalog-policy-go-two-different-tapes
As for why it's just started happening - maybe this is the first opportunity it has had, maybe there are fewer jobs running at the time of the catalog so it has more resources available (e.g. drives) so it can write to two tapes?
03-12-2014 01:29 AM
Is this a problem for you ?
Catalog backup spanning tape volumes is perfectly supported. It was a issue backup in NBU 5.x
03-12-2014 10:09 AM
Thanks for your response. The option to restrict the maximum number of full volumes is actually grayed out.
03-12-2014 10:09 AM
I was thinking of this solution, however, one thing I notice is that the catalogue backup is not using both drives at once. It writes to one tape, than, another job is created and another tape mounted. However, I can give this a shot to see how Netbackup behaves.
03-12-2014 10:11 AM
It's more of an annoyance really. Having to check out two tapes instead of one, or, required two tapes in the event of a catalog restore is a hassle. If my catalog was larger than 1 TB, I could understand this behaviour and would accept it.
03-12-2014 03:39 PM
I think what you are saying is that the first component of the Catalog backup, being the Postgress DB backup job writes to one tape; then the catalog file backup is selecting and writing to another tape?
Strange; normally the tape would still be online to the Master server and ready for use so the catlog file backup would just allocate the same media, re-position and start writing. Certainly that's what happens in my case; I even create a second copy.
The only time a new tape is used is when the first fills up.
Has anything changed? Was v7.6.0.1 before; no upgrade performed?
03-12-2014 04:58 PM
I think what your saying is exactly what is happening. This is a fresh install of 7.6.0.1 on a windows 2008 R2 box. No changes have been made as far as the policy or STU are concerned. I can't think of anything that would cause this. When I get back from vacation, I'll post some logs in this thread to see if we can get to the bottom of this.
Thanks for your response.
Seth
03-13-2014 12:27 AM
Should've posted this yesterday when I found it, altho' I did think that maybe Mark would've popped in directly himself!
You could maybe try Marks solution from this post?
https://www-secure.symantec.com/connect/forums/can-single-catalog-policy-go-two-different-tapes#comment-7673701
There is a setting in the volume pools that restricts the maximum number of partially full media - but this is disabled for the catalog pool even though it would be the perfect solution provided you leave the catalog tape in the library until it is full.
However they didn't stop it being set via the command line:
vmpool -update -pn CatalogBackup -mpf 1
- one potential caveat though is the comment "...it would be the perfect solution provided you leave the catalog tape in the library until it is full."
03-13-2014 01:19 AM
If you want to go simple in a event of catalog restore backup to a NFS/CIFS share configured as a basic disk storage unit. This is the most simple way to restore the catalog without the inconvenience of re-creating media servers, storage units, tape drive etc etc.
This is how I configured my most critical Netbackup domains (and restores is tested as well).
03-13-2014 03:33 AM
I dont think this is a 7.6 thing as it has been known to happen in other environments as per the link Andy Welurn supplied earlier which reflected exactly the same thing.
It generally comes down to tape drive load balancing which is built into NetBackup
NetBackup tries to balance the number of mounts each drives has so if you fire off two jobs, one after the other, the first will go to the first drive and the second will go to the second drive just to keep number of mounts on each drive the same.
What actaully happens is often down to pure luck .. if while the first part of the catalog backup is running the other drives get plenty of mounts then it will carry on using the same drive, if not it will swap.
It all seems very random but there is some sort of logic behind it somewhere and the only way i have seen to make sure it doesn't happen at some point is to use the command as per the linked thread to allow only one piece of partially full media - though you may actually see it be moved from the first drive and mounted in the second drive if NetBackup thinks that is best.
Hope this helps, let me know if:
vmpool -update -pn CatalogBackup -mpf 1
still works in 7.6
03-22-2014 10:02 AM
I tried this solution, but, the catalog backup still is using two tapes. It was a good suggestion though !
03-22-2014 10:03 AM
I'll input this monday and get back to you. Thanks!
03-24-2014 10:47 AM
vmpool -update -pn CatalogBackup -mpf 1
This seemed to have resolved the issue for now! Thanks!