Backup plan for 3-5 years

[Asder]

Hi,
I'm new to veeam backup software, just hope somebody can give me some advice or plan.

Now moving from other big beckup software C. I have to store lot of backups for years...
I have to make full backup of multiple vmware virtual machines once a week and incremental all other 6 days. Store them for 1 month on disks and after that on tapes.
ButI don't have so much space on storage to have non-deduped 5 fulls and 25 increments.
I can't understand what configuration to use without global dedup in veeam. 4-5full backups which are not deduped for whole month are very "heavy"...
Tried to use windows server deduplication but without success.. Now I'm going to disable compressing in backups, hope it will make windows dedup able to work.
What type of full backups is better to use in such case?If I use synt fulls and write everything to tapes - will it be possible to restore backup for ANY day, for example 3 years ago?

[HannesK]

Hello,
and welcome to the forums.

For customers who are new with Veeam, I always recommend the quick start guide

I can't understand what configuration to use without global dedup in veeam

XFS or REFS file system with synthetic full backups

Windows deduplication makes no sense, agree (there is actually a warning if somebody is using NTFS instead of REFS)

If I use synt fulls and write everything to tapes - will it be possible to restore backup for ANY day, for example 3 years ago?

if you have enough tapes, then yes I believe, most customers don't do such things, but from a product perspective it's possible. Just define the "protect data for" 36 months (GFS media pools daily retention only allows 999 days, that's why I suggest a standard media pool)

Best regards,
Hannes

[Asder]

OK, thank you, I'll have to try synt fulls in another datacenter next week. Now I configured active fulls as in previous software, but space consuming is not good..
Active full seems to be more stable even synt... What if some software will recover in inconsistent state after merging to synt full? As I understand it has a little chance to happen.

I'm reading guides and blogs&forums, but didn't find anything about such long backups.. Official support channels never can tell full truth if some option isn't anouth stable for such long retention period. They even may not have such statistics..
So I had to ask here for advice. More chances to get real information.

if you have enough tapes, then yes

I think everyone understands, it is not my idea to have backups for every day for years. =))
So we have to use tapes..We don't have choise, as for me...

p.s. - now reconfigured backups as in this post https://www.veeam.com/blog/data-dedupli ... veeam.html - windows deduplications now started saving space. But on start veeam's compression gave better results.

[HannesK]

Hello,
There is always a chance for hardware defects / bit rot / whatever issue. That's why implementing the [url=https://www.veeam.com/blog/321-backup-rule.html3-2-1 rule[/url] is important. With tape, you have a safety net. But you can always add more copies.

Customers usually only post if things go wrong. I know customers with 7 years retention on REFS, but they are not there yet (started 3 years ago or so).

I was never a fan of Windows deduplication. I have seen too much data corruption in Server 2012...

Best regards,
Hannes

[Asder]

I was never a fan of Windows deduplication.

But what choise I have? Without global dedup function in Veeam - I have to find another way to dedup 4-5 different full backups on one storage repository..Without deduplication storage system.
I understand, that in this case risk of data corruption is 4-5x bigger than having 5 copies on storage, but I have no choise really.

[HannesK]

But what choise I have?

REFS on Windows, XFS on Linux, use of synthetic fulls. That's what the majority of customers does.

only very few people use file system based deduplication.

I understand, that in this case risk of data corruption is 4-5x bigger than having 5 copies on storage,

the source of that calculation would be interesting. whether deduplication is done by Veeam, REFS (block cloning) or NTFS has the same result: if a deduplicated block is broken, then it corrupts all files that rely on that block. That's why I mentioned the 3-2-1 rule.

[Asder]

REFS on Windows, XFS on Linux, use of synthetic fulls. That's what the majority of customers does.

Yeah. I understand, that using synthetyc fulls will save space. I'll try to use it at another datacenter next week, as I mentioned allready. But I don't know if it will be accepted by my boss...
Now I'm trying to save space using active fulls, because it seems to be more stable as I and my colleagues think .. We have a lot of self-made software by our company developers, and nobody wants to have risks of inconsistent backups after synt-fulls ...

the source of that calculation would be interesting.

I mean that if I have 5 similar files on storage I possibly have 5 chances to restore it from different backups. And when they are deduped into 1 file - only 1 chance (now I'm not speaking about raids, file system only)

[HannesK]

for 5 different files on 5 different media, the calculation makes sense to me. for 5 vs 1 file on the same media, the risks are different, because if a drive breaks, it often breaks in multiple sections, or the controller breaks and makes the whole drive unusable (I nave no idea how to calculate that risk).

In the end it's your decision: dedupe or not. Veeam is fine with both. XFS / REFS is recommended and if you think Windows deduplication is better more secure than REFS deduplication, then I will not argue that. I just have bad experience with Windows dedupe

[MarkBoothmaa]

Beware windows dedup can cause issues when the optimization job is running with locking files in use. That's why I moved away from an inherited solution that used windows dedup and went with REFS.

[Asder]

In the end it's your decision: dedupe or not. Veeam is fine with both. XFS / REFS is recommended and if you think Windows deduplication is better more secure than REFS deduplication, then I will not argue that. I just have bad experience with Windows dedupe

I don't understand one thing - What benefits from REFS without windedup with Active Fulls(Not synthetic)??
Volume with backups now is formatted to REFS, and active fulls are on REFS now, with windedup running on this volume. Tomorrow will be next full backup, so I'll see win dedup results with 2 active fulls on refs.
If they will be bad - I have no other way - I'll have to move to synthetic fulls.

[HannesK]

What benefits from REFS without windedup with Active Fulls(Not synthetic)??

none. But I rarely know customers that do active full (simply because active full costs compute, networking & IO resources)

[Asder]

Unfortunately, results are very bad.
7.85TB>7.2TB
17.64TB>15.85TB

I see 2 active fulls for each vm in one repository, but dedup can't see it. ¯\_(ツ)_/¯
All compression in veeam is disabled. Going to move to synt fulls, no other way...

[ChriFue]

Hello

We have several installations running Windows Srv 2019 Repos with ReFS and verly long retention.

Daily incremental with weekly synthetic fulls, backup health check enabled.

Oldest restore point is older than two years now, works without issues.

For safety we are still doing monthly synthetic fulls to tape from this repository. So far we did not need them.

We see similar results on Linux/XFS repositories with immutability enabled.

By using enterprise RAID-controllers (Raid6 or 60) with cheaper enterprise (NL-SAS) disks we never had an issue.

We stopped using windows dedup or dedup appliances a long time ago as FastClone with ReFS and XFS works great if you configure your jobs right.
Change rate of the VMs is about 20% a week, perfect candidates for FastClone space savings.

Eventually you need to size your repository a little bit larger, but it is worth it for not having troubles with dedup-black-box magical thinks happening on restores ...

And hey, with VEEAM v12 you will be able to move your data to a new repository without loosing space savings!!

Christopher

[Asder]

FastClone with ReFS and XFS works great if you configure your jobs right.

Hi, thanks for advice.
What should I know about configuring such jobs?)

[ChriFue]

Get sure to have "Use per-machine backup files" enabled and "Align backup file data blocks" on your backup repository.
Formatting ReFS volume always with 64k record size.
https://bp.veeam.com/vbr/3_Build_struct ... block.html

For the backup jobs we go with "Compression level: optimal" and "Storage optimization: Local target"
Gives you very good backup speed with good compression -> short backup windows.
We also use "Perform backup files health check" one a month.
"Create synthetic full backup" once a week on the day you like.

Works well in most common cases for the primary backup, some additional tuning may be needed in special environments.

[Jan2]

Hi,

I have to make full backup of multiple vmware virtual machines once a week and incremental all other 6 days. Store them for 1 month on disks and after that on tapes.
ButI don't have so much space on storage to have non-deduped 5 fulls and 25 increments.

You have encountered the space consumption issue with Veeam. If you use your tape differently, you can save a lot of space. Veeam permits you to backup the latest chain of backups to tape. Consequently, you can perform a continuous synthetic backup of your VMs and be able to roll-back 25 days. Then weekly, you backup the latest chain of backups to tape and store them either as weekly or monthly backups for your rotation. This gives you a tape copy of a full backup and a week of incrementals. It takes a few more tapes to complete it, but it saves on disk space for the VM backups. You can add a monthly full backup of all 25 increments and the synthetic full if you wish. I run this type of configuration. each weekly and the monthly have their own job definition so they rotate tapes properly. If you choose to use this method, you will have a fifth week tape backup that you will have to disable most months. Some moths have one extra weekend so I enable this job during those months.