The need for Active Full backups

[d.lansinklesscher]

I don't know where I read it, but is it true that a copy-job also protects you against bitrot because the copy job compares it's data to the data from the backup Job?

Can I safely run a forever-incremental job(Retention 7) in combination with a copy-job to other storage without making use of sure-back-up>?

[PTide]

Hi,

Yes, I think you can.

[dellock6]

Totally true, and it's the main advantage of backup copy job compared to storage replication. With the latter, any corrupted block is immediately replicated to the secondary backup storage, thus corrupting all the restore copies. Backup copy job read and checks each block from primary backup before sending it to the secondary location, thus any corruption can be immediately spotted.

[gingerdazza]

Hi. We have 2 VMware 5.x clusters (one in each DC), 1Gbps NIC, HP EVA SAN in each DC over 4GB FC. B&R servers in each DC, backing up local VMs- physical DL380 Gen9 LFF with local SATA disks. I want to backup locally in each DC, then run backup copy jobs to each other. I want my restore times to be optimal.

Q1. Why is forever incremental recommended for SureBackup? Will it work with other types?
Q2. With the above setup is it recommended to use forever incremental or incremental with weekly synthetic? And why? (assume I don't want to use SureBackup)
Q3. if I run weekly synthetic on a Saturday, AND ask it to run an active full backup on every last Saturday, will it run both processes or work out that only the active needs to run?
Q4. if I'm running a restore of a VM using a 6 day old full, plus 5 other incrementals, does that have a huge restore penalty? And is this why someone might prefer reverse incremental?

Thanks

[Shestakov]

Hi,
A1. Surebackup will work with any backup method. Forever incremental let you have just one Full backup file and save the storage space. With Surebackup you can make active fulls much seldom. Please check the first page of the topic.
A2. If you don`t use Surebackup it`s recommended to make active full runs from time to time.
A3. It will run Active Full only.
A4. Yes, Reverse incremental provides the fastest restore, but for 6-point long chains restore penalty will not be crucial. You can make some test to see the real results in your infrastructure.
Thanks!

[gingerdazza]

Thanks - great info

[owenwalker.bc]

We're doing a clean upgrade from v5 to v9 this week and looking at using the reverse incremental backup for backup to disk with a 30 day retention period. We will then be backing up to tape the latest backup overnight, with a daily, weekly, monthly tape schedule inline with how we currently do this

Now where I'm not sure is best practise with running active full backups. I've seen information stating it's good to run a active full backup every 1-3 months when using reverse incremental, also seen information stating you don't need to run active full backups with later Veeam versions.

I'm also not sure exactly how the active full works. Does it create a completely seperate full backup, thereby requiring the full size of the VM x2. Or does it work by creating the incremental changes for the previous night. And then doing a fresh overwrite of the latest reverse backup?

Thanks for any help,
Owen

[Shestakov]

Hello Owen and welcome to the community!

Now where I'm not sure is best practise with running active full backups. I've seen information stating it's good to run a active full backup every 1-3 months when using reverse incremental, also seen information stating you don't need to run active full backups with later Veeam versions.

Please review the thread for the comprehensive explanation.
Do you use Surebackup?

I'm also not sure exactly how the active full works. Does it create a completely seperate full backup, thereby requiring the full size of the VM x2. Or does it work by creating the incremental changes for the previous night. And then doing a fresh overwrite of the latest reverse backup?

It creates a newly VBK file from the scratch. No incremental changes used. It requires 2x of the Full backup file, not VM size. Backup size is much lower than VM size due to compression.
Thanks!

[owenwalker.bc]

Do you use Surebackup?

No we have a standard license.

It creates a newly VBK file from the scratch. No incremental changes used. It requires 2x of the Full backup file, not VM size. Backup size is much lower than VM size due to compression.
Thanks!

Ok yes that is what I mean. It requires the full size of a single days VM backup size. As obviously the reverse incremental will have the incremental stuff from the past 30 days.

[Shestakov]

Having no recoverability tests (surebackup) I would suggest to enable Health Check for Backup Files and perform active full backups once in 3 months or so.

[Crof]

Hello,

I'm using reverse incrementals for backups, but I'm looking to have the backup 'refresh' or 'recreate' its base image from the source once in a while to make sure there isn't any corruption. It scares me a bit to think that without something like this running once a week/month that some data could be left 'unchecked' for years.

My first reaction was to check "Create active full backups periodically" thinking that would do what I wanted: no, that creates full backups in place of a normal incremental (Looking back, I'm not sure why I thought it would do anything different).

Will "Perform backup files health check" do what I'm looking for? Or perhaps, "Full backup file maintenance"? Something else?

Thanks in advance!

[foggy]

You can review this thread for some considerations regarding this matter. Basically, with health check enabled and periodic recoverability tests (SureBackup), there's no need for active fulls.

[pirx]

Hi,

we are planning our Veeam setup for a remote location.

- 25 TB data, change rate 5-10%
- 14 dailies, 10 weeklies
- backups must be stored twice (daily at least 14 days, and all 10 weeklies)
- 2 data center rooms
- IBM StoreWiz primary storage (stretch cluster)
- 2x NetApp FAS for backups (NAS), one in each room

The easiest thing seems to be a forever forward chain with 70 restore points. This would result in 105 TB primary backup (10%) and also 105 TB for the BCJ. Does that make sense at all? Is is good practice to create such a long chain without active or synthetic fulls? How big is the impact for restores of such a long chain? We don't plan to use sure backup. Forever forward has the advantage that snapshots can be removed earlier, which is important to use, because at the moment there is no storage snapshot integration for IBM.

[Shestakov]

Hi pirx,
please read the thread for the explanation of full backups vs. surebackup best practices.
I would emphasize that it`s recommended to keep a short chain on the primary repository for faster restores and longer chain on the secondary repository for restores of historical data.
Thanks!

[ferrus]

I would emphasize that it`s recommended to keep a short chain on the primary repository for faster restores and longer chain on the secondary repository for restores of historical data.

Damn - We're doing the opposite of that. We have a 30 restore point Forever Incremental strategy on Primary Fast Storage, and 7 restore points + 4x Weekly & 24x Monthly GFS on a Data Domain.

I hadn't read this thread before - but just noticed this ...

is it true that a copy-job also protects you against bitrot, because the copy job compares it's data to the data from the backup job?

Totally true, and it's the main advantage of backup copy job compared to storage replication

Does that mean, if I have Backup Copy's from the primary storage to the Data Domain - and the Data Domain organizes the Restore Points into GFS backups - I can switch off Health Checks on the Data Domain Backup Copys?

[Shestakov]

Damn - We're doing the opposite of that. We have a 30 restore point Forever Incremental strategy on Primary Fast Storage, and 7 restore points + 4x Weekly & 24x Monthly GFS on a Data Domain.

I would recommend 7-14 restore points on the source site and 30 restore points + 24x Monthly GFS on a Data Domain.

Does that mean, if I have Backup Copy's from the primary storage to the Data Domain - and the Data Domain organizes the Restore Points into GFS backups - I can switch off Health Checks on the Data Domain Backup Copys?

Health check and Surebackup are different things. Health check is a part of backup copy job and going to help with data integrity.

[lando_uk]

I once had a corruption on primary, just one VM in the job had errors, the rest were fine and the only way I find it was a restore test (or surebackup ) or to run the cmd " Veeam.Backup.Validator" against the job. The Copy Job didn't spot this corruption and happily copied it offsite.

For your reassurance, the VBK didn't just go `bad` for no reason, we had a dead HDD in the RAID6 causing the whole server to stop and need a restart to fix, the VBK corruption happened after that event. We now run a Veeam.Backup.Validator task on all jobs when a HDD dies as a precaution, but it can sometimes take a week to check everything.

[ferrus]

I was hoping for a different answer
Have a VM that's takes 13 days for a Health Check on a Data Domain. Desperately trying to find a way round it - 'cos that's half the month's backup copies lost.

Just wondered whether the combination of checking during the backup copy, and merging into full GFS backups would negate the Health Check requirement.

[lando_uk]

This is my biggest worry. I have to tell the business we are 95% sure the backups are good, but there's not enough time to check if they are all good.

To increase the 95%, I'm trying to convince them to invest in a 2016 S2D ReFS cluster with self healing, but even that cant be 100% error resistant.

[ferrus]

Sounds good. With a lot of space consolidation and temporary backup repo migration, I think I can migrate to 2016 - but it still leaves us in the same situation with the Data Domain.

[lando_uk]

I wonder if ReFS 3.1 also improves the Veeam.Backup.Validator or Health Check performance, maybe there will be enough time in the day to check everything?

[mkretzer]

Is Health Check even necessarry with integrity streams?

[Gostev]

Only on simple ReFS volumes (for its backup healing functionality).

[dellock6]

As far as I know you can enable mirror and parity modes also on single windows 2016 servers, without storage spaced direct.

[pirx]

Hi pirx,
please read the thread for the explanation of full backups vs. surebackup best practices.
I would emphasize that it`s recommended to keep a short chain on the primary repository for faster restores and longer chain on the secondary repository for restores of historical data.
Thanks!

But what would be the "best" way to archive the goal to have each backup (at least 10 weekly) 2 times for 10 weeks, one on NetApp A, the other on NetApp B? If we have a primary chain with 14 restore points on NetApp A and a BJC with 10 weeklies on NetApp B, we miss this goal. Do we need a second BJC that also performs a copy of the primary backup job to the same NetApp A? This sounds not very backup space friendly.

[Shestakov]

If you want 10 weekly backups on both primary and secondary repository, the best way indeed is 2 backup copy jobs. I would still set 7-14 restore points on the source backup job and 2 simple + 10 weekly GFS on the backup copy job pointing to the primary storage. Note that you need to create a separate folder on the primary repository and add it as another repo since it`s prohibited for backup copy job to set same repository as source and target.
Thanks!

[iburnell]

Is a Synthetic Full good with daily incremental good enough to ensure integrity?
- Should we be creating periodic active full
- Should I tick box to do Health Check
- Is reverse incremental an idea i.e. to keep the full to always be the latest

Thanks

[Shestakov]

Hi Ian,
please review the answers above. I would recommend to start from this one.
Thanks

[iburnell]

Our Disk backup job run daily Forward Incremental with Synthetic Full on a Friday. I want to ensure the data is good. Questions
1. Given the synthetic full backup is always "synthetic" is it better practice to do periodic active full backups. Is this doing a brand new full from VSphere rather than chaining the .VIB files?
2. What about the maintenance options. Is it a good idea to tick box to perform backup health checks. Also how about defragment and compact full backup

Thanks

[DGrinev]

Hi Ian,

Please read the existing discussion with the answers you're looking for, starting from the post by the link above. Thanks!