Full and Inkremental to Tape - How to Rotate Tapes

[bjoernwolter]

We use Veeam to back up two of our branch offices to the central repository on our backup server (backup-to-disk). The jobs have a retention period of 14 days and create a synth full every Wednesday.
Now we want to save these backups to our LTO. To ensure we make the most of the tape, we want to set up the backup as follows:
On Wednesdays, a new tape should be used, and the synth full created on Wednesday should be backed up to it. In the following days, the incremental backups should also be written to this same tape. Starting next Wednesday, another tape should be used, and the process should start again. Starting week 4, one of the previous tapes can be used.
In short: No really GFS, just the fulls and increments of one week on a tape, which is overwritten after 4 weeks.

How exactly do I configure MediaSet, TapeJob, and what other settings do I need to make?

[david.domask]

Hi bjoernwolter,

A normal media pool can do this, and you'll probably want the following settings:

- New Media set on Wednesdays
- Protect the tape for 4 weeks

Then create a Tape job targeting this media pool with the source job added, start the schedule after the source job(s) complete, and all should work as you're expecting.

I recommend give our Tape FAQ a read as well as it will help with getting you started with Veeam and Tape.

[jcalvetm]

As we do it:

1 - Backup to SOBR (disk backups)

2 - Tape job configure as secondary target. Tape job is scheduled as "as new backup files appear" and in the tape job configuration/Media Pool/Schedule button you select the days when you want your virtual full backup.

But there is an issue (at least in the version we are using now) or probably I am not doing it right. Virtual fulls on tape only appear if there are NOT periodic fulls on the disk jobs. Virtual fulls on tape only appear if we are using a forever forward incremental without full backups (a chain with only one full and all the incrementals you desire, but only one full). So unless I am doing something wrong, which is possble, or there is a fix in newer versions, If you want to configure virtual fulls on tape you will probably have to stop doing the disk synthetic fulls on wednesdays.

Things to be aware:

- Media set creation days (it seems it is fixed, but I do not know, but if a tape exists with data and a media set, and a new media set number is created (in the same tape, i.e. is "rewritten"), old copies prior to the rewritting can be "not accesible".

- Backup to tape "as new backups appear" only works if backup date is "today". If disk copy finishes at 23:59 and tape copy starts at 00:01, Tape job will not find any copy, because dates do not match. 00:00 makes a clear sheet in backup detection. If you have "missing backups", this may be the reason (We had to configure our copies to start at 00:05, and lose the precious hours between "office close" and midnight)... Take into account that "full Wednesday backup" starts on 00:05 on thursday.

There is a bit of trial and error.

Hope it helps.
(And sorry if I am wrong)

Jordi Calvet

[david.domask]

Hi jcalvetm,

> Virtual Fulls

You are correct, as per the Virtual Full User Guide page, Virtual Fulls are utilized when the source backup chain is Forever Forward Incremental. With your current set-up, the synthetic fulls made by the primary backup will be copied to tape instead of a Virtual Full being created. The purpose of a virtual full is to give a full backup for increments on tape to be linked to -- due to the nature of merging in Forever Forward Incremental chains, without periodic fulls being made on tape to link VIBs to, the backup chain on tape can potentially be "broken" due to the merging of the original backup files, resulting in increments on tape without a full backup to link to.

Virtual fulls solve this problem for Forever Forward Incremental by creating an adhoc full directly on tape -- with period fulls enable in the source job, Virtual Fulls are not required.

> As new backups appear timing issues

I see to recall this was a known issue on earlier versions (want to say v11 and possibly early v12 releases), however, the way tape jobs work means that on the next run, backups in the repository without a copy on tape will be queued for backup during the next session. If you're not seeing this on the current version (12.3.1.1139), please open a Support Case and allow Support a chance to review the job logs and confirm the behavior. Please share your case number once created, thanks!

[jcalvetm]

Thanks for the answer!

Agree.

In our case, customer's requirement is that every tape must be "self sufficient". So every tape needs to contain a full backup. To this customer "the important backup" is tape backup. Disk backup are "just an addition".

We use two sets of five tapes (mo to fri) in alternate weeks, stand alone LTO tape unit -manual tape change-. Daily copy. Mo to thu contain a daily virtual full and Fri contains Fri virtual full plus sat and sun incrementals. To be able to recover the whole system with a single tape is a customer a requirement.

As we suffered an outage in the very moment when full was being merged with oldest incremental (that means whole backup chain lost), my train of though was: I configure a synthetic full in my disk backup chain, so if "something happens" when creating the new full in my forever forward incremental chain, only the last week of backups are lost. But I noticed that if I did that, virtual fulls were not created in daily tapes.

I admit hat i need to install last updates in veeam server (and client updates), but the prospect of a production server restart is making me wait for a good moment to do that.

This is because I had to disable the weekly synthetic full on disk repo. Maybe there is a solution, but I have been not able to find it yet.


Best Regards
Jordi Calvet

[bjoernwolter]

A normal media pool can do this, and you'll probably want the following settings:

- New Media set on Wednesdays
- Protect the tape for 4 weeks

Then create a Tape job targeting this media pool with the source job added, start the schedule after the source job(s) complete, and all should work as you're expecting.

I’ve configured it like this: for our first branch, everything works fine—the backups are stored directly on the HQ Veeam repository, which is connected to the backup server. For our second branch, the setup is different: backups are stored locally on a NAS (for performance reasons). We then run a backup copy to our HQ repository, and from there, we write to tape.

The problem is that with this setup, only incremental backups are written to tape. The main backup job for the second branch (with the on-site NAS) is configured to use synthetic fulls.

If I write directly from the NAS repository to tape, everything works as intended: a full backup followed by incrementals. But this isn't an option due to limited network performance.

What I am doing wrong, or could be done to archieve my goals

[david.domask]

Hi bjoernwolter,

Thank you for the additional details -- regarding the backup copy, is it configured to produce periodic full backups (active or synthetic) as well?

I would recommend open a Support Case and let them check the behavior of the Tape Job with the Backup Copy, as at some point it must place a full backup on tape from the chain. Be sure to include logs for the affected tape job and mention which source backup is affected.

Please share your case number once created. Thanks!