The need for Active Full – REFS

[skumflum]

We are running Veeam 9.5 - REFS and it works very well. Although I am excited about the spaceless synthetic full backup, I’m skeptical about the data integrity after eons of block cloning and only transferring delta from VMware. I’m worried about things like CBT horror stories that Veeam would not detect!

What is the general opinion on this?

If I were to run monthly active full backup how does this affect the spaceless synthetic full. Let’s say that retention policy is 28 restore points daily backups with weekly synthetic full and monthly active full. Will the entire set of synthetic full's still remain “spaceless”?

The usage on the repository will be equivalent to two full + numbers of incremental (counting synthetic full as incremental) – right?

[Gostev]

What is the general opinion on this?

If you want to guarantee recoverability, you should test your backups - period. There is simply no other way - plus we make it easy to do with SureBackup, so there's no excuse not to do this either.

Look, you are currently concerned about one of like hundred things that can prevent your backups from being recoverable - and I can tell you that CBT bugs is very low on the list of reasons for failed restores that we're seeing in support (and with 250000 customers over 10 years, we have a whole lot of statistics to work with here).

[aich365]

What are the advantages of Synthetic Full over Active Full?

Apart from the time difference between Incremental backup and new full backup, the advantage of this seems to be lost in the Synthetic merge process which takes ages on a 3TB File? Typically in excess of 48 hours for a Synthetic and 20 hours for a new Full

[skumflum]

If you want to guarantee recoverability, you should test your backups - period. There is simply no other way - plus we make it easy to do with SureBackup, so there's no excuse not to do this either.

Look, you are currently concerned about one of like hundred things that can prevent your backups from being recoverable - and I can tell you that CBT bugs is very low on the list of reasons for failed restores that we're seeing in support (and with 250000 customers over 10 years, we have a whole lot of statistics to work with here).

Okay, I get your point about restore and I have not looked into SureBackup yet. You don’t see any added value in running active full once a month then?
I got close to a thousand VM’s and have gotten into my mind that it was impossible to run SureBackup in any meaningful way – but maybe not.

http://www.virtualtothecore.com/en/can- ... urebackup/

Would this be an approach?

Btw – Am I correct about the space requirements?

[skumflum]

What are the advantages of Synthetic Full over Active Full?

Apart from the time difference between Incremental backup and new full backup, the advantage of this seems to be lost in the Synthetic merge process which takes ages on a 3TB File? Typically in excess of 48 hours for a Synthetic and 20 hours for a new Full

Not With REFS Fast Clone... I have one job with 20TB of VM's - the synthetic full takes 15 minute.

[Gostev]

You don’t see any added value in running active full once a month then?

The only added value is perceived peace of mind. And I am not being sarcastic here - honestly, if this will make you sleep better - then by all means, do it. One less thing to worry about in your life is worth a lot, at least that's my approach to what troubles me.

All I want to make sure is that you understand this is not a replacement for full recoverability testing. Doing Active Fulls will reduce the chance of failed recovery by may be 1% at best - but you really want to be 100% sure, and only SureBackup can give you that (because what it does is actual restores, just into the isolate lab and without moving the data around).

Ironically, SureBackup loves CBT bugs specifically - in my experience, these either cause OS fail to boot, or server applications fail to start - so they are caught immediately unlike bit rot for example, which is best detected by disk scrubbing with storage-level corruption guard functionality (or the same non-default SureBackup job option).

[skumflum]

Thanks for your directness. Although I do think there is such a thing as a placebo effect, I do prefer to change my mind when presented with the right argument.

[DerOest]

Well, this may be the Internet-Feedback-Effect (I'm sure there's a KXCD for this...): You only provide feedback when you have a problem/complaint, but not when it's working as expected.
The result is that readers (like myself) often see these catastrophic reports (VMware CBT, Microsoft REFS trouble + dedup corruption, etc).

To me, such issues are a huge deal - how would I justify to my boss "Hey, we just lost those 1,5 years of backups because i didn't proactively run active-fulls once a month - sorry".

I totally understand Gostev, he has a much broader view and knows that only less than x% of customers are ever effected by such bugs, but hey, i might be one of those!

So yes, I will always run active fulls periodically. One day it will save my ass

[NightBird]

Keep in mind that active full doesn't resolv CBT bug, because active full use CBT

[ian0x0r]

Following the 3-2-1 rule (3 copies of data, 2 media types, 1 off site) https://www.veeam.com/blog/how-to-follo ... ation.html would help to mitigate some risk, but yes there is no substitute for testing backups.

ReFS fast clone is great technology, just bear in mind the other considerations out there that have been documented on the forums, such as the recommendation to run 64k cluster size when formatting volume ReFS.

Ian

[mkretzer]

Why is there no way to not do an active full but instead do a reset of CBT? This would be much more usefull especially with REFS would it not? Will Veeam compare every block it has in the backup with production storage then?

[Gostev]

Sure there is, it's a simple PowerShell script to reset CBT on all VMs (and we even have KB for this). In which case yes, Veeam indeed will read the entire source image to determine the changes comparing to known latest state in the backup file (based on digests). I emphasize "known" because of bit rot possibility - in other words, digests contain "optimistic" representation of backup files

I've got an idea last year to make manually run Active Full backups to provide an option to automatically reset CBT. Would not want to do this every time, because ability to leverage CBT during Active Full is a big deal for backup window (not having to scan through TBs of empty space).

[mkretzer]

Ok Gostev, but we have storage with integrity bit active so bit rod should be detected on that level. Also, REFS should find bit rod by itself should it not?

So having the option in the GUI to do CBT reset instead OR with active full would be nice.

I know that it works with Powershell but we do scheduling in Veeam and do not want to do scheduling for powershell as well!

[sg_sc]

To jump in on this, what would be the storage consequence if one would do the monthly active full.
The active full would take up the full space, correct?
The next synthetic full after that would use fast clone again, and only use up the changed space, correct?

thanks

[Gostev]

Yep, totally correct.

[couchman75]

Hi
Sorry to bring this thread back from 4 dormant years but it covers my situation well. We moved from 9.5 a few months back where we were using backup copies to keep historical backups on the same REFS volume as our regular backups. This worked OK and we could keep over 6 months with the REFS savings, but of course we were effectively managing 2 backup chains. We then upgraded to v11 and decided to use the new "Keep certain full backups longer for archrival purposes" feature taking the last synthetic full backup of the month, however I'm finding that the savings are nothing like previous and I'm almost out of space after 3 months. We do an active full for the first backup of the month for the same reasons as others on this forum. So I guess my question is, is this impacting REFS and our backup chain? Are the REFS savings being reset even though we are keeping synthetic fulls?
Thanks
Russ..