Tape Daily - Full possible? - Design Questions

[smileyface]

Hi there!

We are using the following setup -

Dedicated Veeam Server with local storage - ReFS - Job 1
Hardened Linux Repo on another dedicated Server - Job 2
Tape - GFS Tape Job 1

Our goal would be to write a full backup to tape every day, so we only need one Tape to restore.
To achieve that, currently Job 1 writes reverse incremental, so the tape always writes full. Now we saw that fastclone only really benefits with Synthetic Fulls - that would "destroy" our tape logic. The questions:

1) Is there any way to always write full backups to tape without reverse incremental source job? I think that daily synthetic full might achieve the same result, but I'm unsure about how much HDD space is used for that.
2) We use LTO8. Our fullbackup is fairly small. Can we somehow "reuse" the tape - so it doesnt get erased before writing to it again? We could easily use a set of tapes for two weeks, there's plenty of space..

Is it that uncommon to write full backups to tape every day? Apparently that's not a standard option, I thought that would be the "nicest" for smaller environments..

[Mildur]

Hi Smileyface

Reverse incremental backups also benefit from FastClone. Instead of copying blocks to a new VRB file, we can just create a new VRB file and reference to the existing data blocks.
Neither less, our recommended backup format is Forward incremental backups with regular full backups for reFS. For hardened repositories regular full backups are mandatory.

1) Is there any way to always write full backups to tape without reverse incremental source job? I think that daily synthetic full might achieve the same result, but I'm unsure about how much HDD space is used for that.

Daily synthetic full backups can be used as a workaround. It won't use more diskspace on reFS or XFS filesystems with our FastClone feature.

2) We use LTO8. Our fullbackup is fairly small. Can we somehow "reuse" the tape - so it doesnt get erased before writing to it again? We could easily use a set of tapes for two weeks, there's plenty of space..

You can append to incomplete tapes as long as you want, but Mediasets cannot be auto closed after 2 weeks.
I recommend to use the same tape not more than one week. Default setting for the daily media is to append backup files till the media set is closed and exported. You will have 6 daily full backups on the same LTO 8tape.

General recommendation:
Tapes should be switched between different tape sessions. Imagine you have 12 days of backups on your tape. On day 13, you have an incident in your datacenter. Your reFS repo, hardened repository and tape drive are affected. No backups left to restore. By switching tapes regularly and bringing them offsite, you make sure to always have a most recent backup copy available to recover from any disaster.

Best,
Fabian

[smileyface]

Hi Mildur,

thank you for your insights.

1) It is not related to tape - but I hope it is OK here aswell - but then our Fastclone is not working properly. Used disk space is 3-4x on ReFS with reverse incremental vs linux repo.. Is our Fastclone not setup correctly, or is the big savings in space only synthetic full backups vs "normal" full backups? In another forum reply here someone stated that synthetic full is recommended for every fastclone repo - is there any "official guideline"?

For hardened repositories regular full backups are mandatory.

Or do you mean synthetic full with this one too?

2) Currently that is not what it is doing for us. Is it because it is a GFS Job with expiration dates? "append backups" is checked in the media set, yet it overwrites the tape if its expired or doesnt append anything if it is not yet expired.

We wanted to remove the tape daily. So we have five tapes - monday through friday. Then on next monday, reuse the monday tape (so two full "monday" backups are on the "monday tape") - same for tuesday and so on. Only the current tape is onsite, the rest is offsite. But this way we could have 10 restore points with 5 tapes - instead of 10..

[Mildur]

Hi Smileyface

1) Difficult to say without knowing your infrastructure.
- Are you doing active full backups with your reverse incremental backups?
- Do you have the exact same amount of restore points on you reFS and XFS repository?
- Does your backup copy job copy all restore point (immediate copy) or does it only once on a schedule (periodic copy)?
Reverse Incremental and Forward Incremental with regular synthetic full backups backup jobs should have a similar amount of restore point and data.

I personally prefer regular synthetic full backups with reFS. I will have smaller backup chains, even if the same data blocks are used.

Or do you mean synthetic full with this one too?

Yes, I meant synthetic full backups.

2) When a media set is closed and exported, an expiration date is set. Expired tapes can be reused by the tape job. Why would you wanna keep an expired tape. If it's expired, then it's free to be used again.

We wanted to remove the tape daily. So we have five tapes - monday through friday. Then on next monday, reuse the monday tape (so two full "monday" backups are on the "monday tape") - same for tuesday and so on. Only the current tape is onsite, the rest is offsite. But this way we could have 10 restore points with 5 tapes - instead of 10.

Tapes have to be closed and exported daily to use a new tape the next day. I cannot see any option how you can achieve your plan. Maybe you can use LTO-7 tapes to save a little bit of money. A LTO-8 drive can write LTO-7 Tapes (please check with your vendor).

Best,
Fabian