Assistance with Backup

[gerdawg]

I'm wondering if someone could help lead me in the best direction for backing up our data. Here's the scenario I'm in and where it's going.

Current scenario: (We run backups 7 days a week)

I'm backing up about 50 VM's of various data to a local NAS box that has an LTO 6 tape drive (2.5TB Native capacity) installed. We are backing up our data every day to disk using reversed incremental and the job has a 30 days retention period. The backup size is about 3.5TB for the full backup and 150gb each incremental for a total of 6.5TB total for the month. The backup to disk seems to be working well for us.

Each week we take the full backup file and put it to tape, this occupies 2 tapes. (Backup to tape job) (3.5TB)
Each month we take a the full backup and all incremental backups (30 increments) plus a few other files that are backed up using Veeam Endpoint backup, this occupies 4 tapes. (Files to Tape job) (9TB give or take)

What we do not have at this point in time is a tape backup of all of data for each day and management is asking that we address this so that each day a backup of our data is stored offsite in the event of a disaster.

I'm assuming I'll have to do away with reversed incremental to meet the demand but I'm not clear the best way to go about this. Specifically, do I create a synthetic full each week and back that up? I still need to retain on disk 30 days of backups preferably but then my data becomes huge and uncontrollable when backing up to tape. (4x3.5TB+incremental to cover monthly). I'm looking for some ideas on how users would go about addressing this need while keeping in mind that we run a backup each and every day due to our environment. Any thoughts are appreciated!

[EugeneK]

Hi Gerry,

Are you looking to address it by maintaining the same infrastructure ie disk for primary and then tape for archived data, or open to offsite backups with the solution like BaaS using Veeam Cloud Connect?
Thanks!
Eugene

[Shestakov]

Hi Gerry,

What we do not have at this point in time is a tape backup of all of data for each day and management is asking that we address this so that each day a backup of our data is stored offsite in the event of a disaster.

I`m also curious what is going to be the offsite destination.

Actually disk+tape protection is a good combination, so one of the options is to copy those daily increments to the tapes(it will be better to switch from reverse to forever forward increments though).
If you want to copy on the disk repository(cloud or your own), you need to use backup copy jobs. I would decrease the lenght of backup chain in the primary repository and have more backups on the secondary one as most of restores happen from the latest restore points.
Here is a good blog post for your case.
Thanks!

[gerdawg]

It will eventually turn into BaaS however we are looking to leverage existing resources currently. Daily offline destination is tape.

[Shestakov]

So you need to switch the backup method to forever incremental, otherwise you will copy full backups daily.

[gerdawg]

Thanks Shestakov,

That's kind of where I was leaning. The part I'm not clear on is how it affects my tape backups. My understanding of the forever incremental is that it is predicated on the initial "full backup". I assume unlike a differential backup, every incremental is required in order to restore the full working data set. So if I were to run a full job on December 1 and I need to backup each additional incremental run to tape every night, on day 28 am I now required to restore 28 tapes in order to get my full backup working in the event of a disaster? What happens on January first? Do customers then generate a synthetic full that backup and continue forever increments until February? I know disk to disk is what Veeam was built around but for now we need tape and there seems to be some significant challenges getting backups to tape. I'm looking for the best methodology to achieve my goals while continuing to keep Veeam as automated as possible since we have other support staff that actual run/monitor the jobs. Can you shed some light on what the best way to get forever incremental to tape, keeping in mind I have daily, weekly and monthly rotation needs?

[EugeneK]

For forever incremental you do need the full chain of Full + incrementals based on it to be present and healthy. If a concern is around the data integrity for any particular data set, synthetic full on weekly or so basis is the option you're looking to get use of. Every next increment will be based on the new full, until the next one is created, in turn, narrowing down the failure domain. The benefit would be not to take the copy of Full backup on daily basis, but rather do so on the times Synthetic or Active Full are created, saving space and lowering the needs for backup window(s).

[Shestakov]

Gerry,
With GFS Media Pool, you can keep historical backups aside of daily ones.
By the way starting from v9.5 there is an option of direct VM restore.

[gerdawg]

For forever incremental you do need the full chain of Full + incrementals based on it to be present and healthy. If a concern is around the data integrity for any particular data set, synthetic full on weekly or so basis is the option you're looking to get use of. Every next increment will be based on the new full, until the next one is created, in turn, narrowing down the failure domain. The benefit would be not to take the copy of Full backup on daily basis, but rather do so on the times Synthetic or Active Full are created, saving space and lowering the needs for backup window(s).

Thanks to both of you for the replies. It's appreciated. The part I'm looking to wrap my head around Eugene is for my monthly backup then. It seems if I do a full synthetic every week then my monthly tape job won't retain the rest of the month since the increments are now rolled into a complete full backup. So if I had to restore my tape for December in January I'd likely only get the last week of December. Does that sound correct?

[gerdawg]

Does anyone have any suggestions on how we accomplish the monthly backups in the above scenario?

[Shestakov]

GFS restore points are all full backups. Synthetic full on the disk doesn`t change the tape fulls so you will have all weekly backups of December in January if you set a proper retention.
By the way, I would suggest to use Surebackup for recoverability checks.
Thanks

[n0creativity]

Have you done a cost analysis on this? LTO 6 tapes are about $30 per last time I checked (maybe cheaper in bulk) but even at $20 per tape, your looking at $600 per month in media costs. Now, if your planning on recycling tapes that's different, but if not you should really look at some Cloud Connect providers.

[gerdawg]

GFS restore points are all full backups. Synthetic full on the disk doesn`t change the tape fulls so you will have all weekly backups of December in January if you set a proper retention.
By the way, I would suggest to use Surebackup for recoverability checks.
Thanks

Shestakov - with GFS restore points I thought that the backups would only be a full copy on tape. This wouldn't be what I'm looking for if so. Can you confirm?

I would prefer to keep backups as such on disk:
30 days with restores throughout the full 30 days. Currently I'm using reversed incremental, so that works for me. (But obviously this will need to change based on offsite tape requirements)

On Tape:
1) A weekly full tape run at the beginning of the week, daily incremental to tape the rest. (I'd ideally like to rotate the daily incremental tapes back into circulation for the next week...we only care about recovering from tape for the night before if needed in a disaster.)
2) A Monthly full tape that includes the 30 days of incremental. (Audit requirements)
3) A Yearly tape job that can just be a full.

I believe GFS would break down on the 2nd line item since I would only capture the full backup on the month end. Additionally I'd like to reuse the Monday-Saturday tapes if we ran a full on Sunday. Is there any better way to do this?

[nunciate]

I run backups to disk every night. Almost all of my jobs are set 28 restore points and to active full once weekly with daily incremental because I have the space to do that. If you are limited on storage I would recommend a full backup followed by daily incremental. Enable create synthetic full once a week (on a weekend day) and keep 30 restore points. If your storage can handle it use the transform previous backup chains into rollbacks (requires a lot of IOPS)/

For the tape jobs simply create the jobs and use the backup job as the source. GFS Media Pools don't support daily backups, only weekly, monthly quarterly or yearly. If you want to keep daily backups offsite for a shorter term than weekly or monthly you will need to use standard media pools. I have 1 Media pool for Daily, Weekly, Monthly and Yearly tape outs. I set my tape jobs to use Daily and Weekly. Once a month I run a script to change my tape jobs to use the Monthly pool and then once a year to use the Yearly pool. Then after the monthly runs I have to set them back to Weekly. It is a pain but I have found no other way to do what I want to do. I use the following offsite retention. Daily for 7 days, Weekly for 1 Month, Monthly's for 3 years and Yearly goes off indefinitely.

[Shestakov]

Shestakov - with GFS restore points I thought that the backups would only be a full copy on tape. This wouldn't be what I'm looking for if so. Can you confirm?

Yes, with GFS there are only full backups.
If you want to store daily backups along with weeklies and monthlies you need 2 Media Pools, one GFS and one regular. GFS will store all full backups while regular Media Pool will have a chain of Full+incrementals.