Trouble setting up right tape strategie

[StephanF]

Hi,

I am setting up the jobs for my backup strategy at the moment and need some advice how I can achieve my goals. Especially the tape part drives me crazy.

What I have at the moment:
1 primary disk repository - target for all backup jobs, source for short term restore job
1 archive disk repository - large and with deduplication, for long time archival of restore points (GFS retention)

I created several backup jobs for my VMs.
- Some do 1 backup a day, some do backups several times a day
- They all are targeted to the primary disk repository
- backup mode is Incremental with an active full once a week
- each job has between 14 and 60 retention point (leading to about 2 weeks retention)

What I want to achieve:
Short term retention is fine with the settings above. But I want to have more safety for the short term data by doing copies and have the ability to do restores of older data by using GFS rotation scheme.

On disk think about the following GFS scheme:
8 weekly (on Sunday)
12 monthly (on 1st day of month)
5 yearly (on 1st day of year)

I think I am able to achieve this by creating a backup copy job from the primary to the archive disk repository and choosing the right retention settings. My tests look good so far here.

On tape I want to have 2 thinks:
- copies of the GFS restore points mentioned above (same retention period as on disk)
- daily one full backup to tape of the last backed up state of all VMs (minimum retention: some days, maybe 3-5)

How do I have to set up the tape jobs to achieve this (or something near this)?

Thanks
Stephan

[Shestakov]

Hi Stephan,

You plan looks sane in general. Here are some comments:

backup mode is Incremental with an active full once a week

It`s not necessary to do active fulls that often. Once a month is more than enough.

Backup copy GFS retention should, but I`d suggest to schedule monthly jobs on "First Sunday of month" to save more space on the repository.

Regarding tapes,
For now there is no native GFS support on tapes(working on it), but there is good workaround:

I would recommended to create 3 media pools and 3 backup to tape jobs writing backups there as described above in the topic:
Create a media pool called “Weekly” with retention period equal to 8 weeks.
Create a backup to tape job called “Weekly” and point to the “Weekly” media pool and schedule it to run on weekly basis.

Create a media pool called “Monthly” with retention period equal to 12 months.
Create a backup to tape job called “Monthly” and point to the “Monthly” media pool and schedule it to run on monthly basis.

Create a media pool called “Yearly” with retention period equal to 5 years.
Create a backup to tape job called “Yearly” and point to the “Yearly” media pool and schedule it to run January 1st.

daily one full backup to tape of the last backed up state of all VMs (minimum retention: some days, maybe 3-5)

Do you need a full backup every day and to keep 3-5 most recent full backups? If yes, you can either switch backup jobs to reverse incremental mode and make tape job run daily or work with Virtual Full Backup.
To copy backups of all VMs on tapes choose primary or archive repository as a source for the backup to tape job and it will copy to tapes all newly created restore points.

Thanks!

[StephanF]

Hi Shestakov,

thanks for your detailed answer. In general this looks like they way I planned it. Just let me verify a few things.

It`s not necessary to do active fulls that often. Once a month is more than enough.

I know about this fact. I think doing active full backups "just give a good feeling" instead of beeing really necessary. I will look to do them less frequent and maybe switch to incremental forever after setting up some SureBackup jobs.

Backup copy GFS retention should, but I`d suggest to schedule monthly jobs on "First Sunday of month" to save more space on the repository.

Can you give me the point why "First Sunday of month" saves more space on the repository than "First day of month"? Is it because the first sunday matches one weekly GFS restore point and so only one point is created? This makes sense.
I will think about this though I am not scared about space requirements thanks to large disks and deduplication.

For now there is no native GFS support on tapes(working on it), but there is good workaround:

Looks good and I already tested something like that. I Think this works good until real GFS on tape will be available.

Do you need a full backup every day and to keep 3-5 most recent full backups? If yes, you can either switch backup jobs to reverse incremental mode and make tape job run daily or work with Virtual Full Backup.

So that's the point we have to talk about a little bit. I am very open for any good advice here. Let me explain why I thought about daily fulls to tape with a short retention:

Both disk repositories I spoke about are in the same location together with the production storage (same location, not same storage!). There is no second location to use at the moment and cloud backup is something to maybe come in the future. So I need a second copy of the last backups daily on tape to get them stored in a safe location as fast as possible to be prepared for a disaster. Additionally I don't think incrementals on tape (especially long chains) are a good thing to do because of the risk of one tape failing. Incrementals on disk are fine.
So the only solution left I can imagine is doing full backups to tape daily. The most recent 3-5 Days are ok because in case of a disaster I only need the latest.

If I understand correctly there are the following ways to do this:
- Virtual Full Backup - This is the most charming method but requires Incremental forever jobs on disks, which I don't have yet.
- Reverse incremental backup mode - This always creates a full backup as the last retention point which I can the backup to tape
- Incremental backup mode and Synthetic Backups every day - This should bring nearly the same result as Reverse incremental backup mode

Any advices of what would be the best or you I can redesign this for best safety in case of a disaster?

[Shestakov]

I know about this fact. I think doing active full backups "just give a good feeling" instead of beeing really necessary. I will look to do them less frequent and maybe switch to incremental forever after setting up some SureBackup jobs.

Up to you, if you are not short in resources, you can make it every week.

Can you give me the point why "First Sunday of month" saves more space on the repository than "First day of month"? Is it because the first sunday matches one weekly GFS restore point and so only one point is created? This makes sense.
I will think about this though I am not scared about space requirements thanks to large disks and deduplication.

Correct. Since you have weekly retention = 8, you will save space of 2 .vbk files.

I don't think incrementals on tape (especially long chains) are a good thing to do because of the risk of one tape failing.

In fact, there is nothing to be afraid of about increments on tapes.

I would suggest 2 plans and you can choose one you like more.

1. Not to be afraid of increments on tapes and leave backup jobs as they are, create GFS-tape jobs as was described + backup-to-tape job running daily with full on Sunday, retention(protection period) = 7.
2. If you need those full backups on tapes, switch backup jobs to reverse incremental method and create a backup to tape job with desired retention(protection period). Also create GFS-tape jobs as was described.

Thanks!

[StephanF]

In fact, there is nothing to be afraid of about increments on tapes.

Not on the side of Veeam. But if I have to change the tape daily to put it in a safe place I end up with a chain over many tapes. One failing tape will destroy the whole chain. This is why fulls (or maybe differentals) make sense on tape but not incrementals. Just my two cent.

I would suggest 2 plans and you can choose one you like more.

1. Not to be afraid of increments on tapes and leave backup jobs as they are, create GFS-tape jobs as was described + backup-to-tape job running daily with full on Sunday, retention(protection period) = 7.
2. If you need those full backups on tapes, switch backup jobs to reverse incremental method and create a backup to tape job with desired retention(protection period). Also create GFS-tape jobs as was described.

I think I'll go with the reverse incremental mode for now.
By the way, this scenario is why I suggested to allow Virtual full to tape also for incremental jobs with active fulls in this thread.
So you can have the best of two worlds.

[Shestakov]

I`ve read the linked post. Great feedback, we really appreciate it, keep sharing your thoughts!
Once you need an assistance with the backup strategy, let me know!

[Dima P.]

StephanF,

By the way, this scenario is why I suggested to allow Virtual full to tape also for incremental jobs with active fulls in this thread.

And we got this request added to the list. However, this is not going to be included in the upcoming major release – more likely in the next verion