Move to GFS Tape Jobs

[nunciate]

Well after many months of tinkering and testing I am now ready to migrate to GFS tape jobs. I'd like to get a little input before doing so.

Today I have 34 backup to disk jobs in production with about 200 VM in them. Backups start around 5:00 PM and run through the night.
I have another 34 regular tape jobs that correspond to the backup jobs. Those start at 10:00 PM each night and also run through the night.
Currently, some tape jobs fail because backup jobs are still running but typically get caught the next day or the next evening so no big deal there.

My question is this. If I move to GFS should I consolidate all my tape jobs? I have a tape library with 5 tape drives.
Should I consider just running 5 GFS jobs and spreading out all my backup jobs within those or should I continue to have 34 tape jobs to match each backup job?
The current tape job schedule is set to start 5 jobs at 10:00 PM and then all other jobs are chained off of those so that only 1 job is using 1 tape drive at a time.

What are all of you doing?

[HannesK]

Hello,
I would consolidate a bit. 34 jobs for 200 machines would be too many for me personally. Something between 1 and 5 GFS jobs is something I would do.

5 GFS jobs

My feeling is, that the "5" comes from the number of tape drives. I would use parallel streaming to multiple drives and make the number of jobs independent from the number of jobs. Avoiding job chaining is also a good idea from a complexity and "error prone" perspective

Best regards,
Hannes
PS: I assume that the backup jobs are pointing to repositories with per-machine backup chains (best practice and requirement for parallel streaming of one job to multiple tapes)

[nunciate]

Yup, I am using per-machine backups. I'll go ahead and setup like you recommended and see how it goes. Thanks for the advice.

[nunciate]

I figured I would post a follow-up on how this went for me.

I modified all of my back-ups to disk jobs so they are all set to Incremental (Recommended) and Synthetic Full is not checked. I set the retention to 30 restore points.
I have about 40 jobs for my VMs in Prod and about the same for DR.
I let those run for 30 days so the old synthetic full backups all ran off and that went quite well. My storage cleared up considerably over time and my backups run much faster now that I don't have to wait for weekly rollups and such.

I then removed all of my tape jobs. Previously I had 1 tape job for each of my backup jobs.
I replaced them all with 5 jobs in Prod and 5 in DR. I chose 5 jobs as we balance our full tape out over a 5-day period. So each tape job contains servers that do full tape out on those days.

So far so good. I have been successfully doing full backups on the days they are supposed to go and incremental to tape every other day in the week.

I am very please with the speed and overall performance of the jobs.

[HannesK]

thanks for the feedback, good to hear