Same GFS backup to two tape libraries

[479er]

Hey everyone,

I'm having a few issues implementing GFS to Tape. I have two tape libraries, where I have to store GFS monthly backups. Both libraries must store the same monthly backup (tapes on one library are taken off-site, and the others stay on-site).

At the moment, all VMware backup jobs create an Active Full backup on the 1st Saturday of each month (monthly GFS enabled for first week), and I wish to take this monthly backup to tapes in both libraries.
So, I made two GFS media pools, one for each library, and two GFS to tape jobs (targeted at each library) that run the same day when the source jobs create an Active Full backup.

This is extremely slow, as both libraries are trying to read from the same source at the same time. Moreso when one has to create a synthetic Full whenever the source backup job fails to create an active Full.

I'd rather not make a separate backup job (scheduled monthly) and take them to standard media pools in each library, as that was our old setup. If there's no better way of doing this, then I may have to failback to that setup.


I would appreciate any suggestions or comments on the matter.

[eblack]

Have you looked into tape parallel processing yet?

[479er]

Hey @eblack,

I'm not sure I understand. The libraries are working in parallel, but that is exactly the cause of slow performance using GFS to tape from one source to two libraries at the same time.
Hope this clears up what my issue is.

Regards,

[david.domask]

Hi 479er,

Tape Copy may help here, but there are a few nuances to consider which might make it a non-starter for you:

1. Currently encrypted tapes/backups not supported
2. When a copy is performed, the Tape Catalog will show the restore points from the target of the tape copy (i.e., if you write backups to Tape A ,then copy Tape A to Tape B, Tape B is what the UI will show as holding the backups)

I think 2 is manageable, as you can simply do the actual Tape GFS backup to the library that has its tapes off-site'd, then copy to the library where tapes will stay on-premises, but with off-siting the tapes, I understand 1 may be a challenge.

Alternatively, consider changing the start time of the GFS scan period in the job setting so they don't overlap -- if you have a relatively predictable tape-out schedule, this should work as well.

[479er]

David,

Thanks for you input. Tape Copy isn't really an option because I don't want this to become a manual task.
I guess I can play with the start time of the GFS scan period, as you said, but I don't think it will produce the results that I'm expecting.

[vmtech123]

I have 2 libraries as well and have to create two pools, 2 jobs etc.. But I have a repo at each site...

Backup to site a, copy job to site B repo... then each tape job runs from their respected repo.

where 2 tape libraries is a real pain is file to tape jobs, as it uses double the VULS

[479er]

Hey vmtech123,

Having a repository in each site is what I strive to have but, unfortunately, we do not have the budget at the moment.
For the time being, I have enabled synthetic fulls only, that way the main backup jobs finish before the GFS to tape scheduled time.

Thanks everyone!

[vmtech123]

Schedule is important to not have jobs locked, Tape jobs can tend to take a long time in some cases. Many of my VM's are 30-+TB

With only 1 repo, and 2 libraries, Create a backup job, then two tape jobs on different schedules. Tape jobs REALLY hammer the SAN if it's spinning disk depending on how many parallel tasks you are running. It's a ton of data, and if you are using REFS like me that data is not sequentially stored. It makes sense that if you ran both at once on similar backup jobs you will have high latency or swamp the drives.