Any reason to NOT go Server 2022 on new backup server?

[pufferdude]

I will soon be installing a new Veeam B&R server (doing Veeam db/tape/primary repo duties) with 12x8TB DAS (+ mirrored SSDs for OS) to replace our ancient Dell R520 server doing the same tasks today.

Is there any reason (Veeam compatibility in particular) to not put Win Server 2022 on this new box?

[Gostev]

No reason not too. No doubt this is what I would do, particularly due to the ReFS metadata engine optimizations it brings (previously available only in SAC version 2004).

[Regnor]

Please only be aware that the bundled SQL Express 2016 which comes with the Veeam ISO isn't supported on 2022. So you'll either have to manually install a newer version or upgrade it after the installation.

[Gostev]

From Veeam perspective it's not a problem. We agreed to support such installations normally because we have not seen any issues during our own Server 2022 support testing. So unless you're specifically concerned about an ability to open an official support case with Microsoft in regards to the database issues, which is quite unlikely when talking about free SQL Express in the first place, then you can forget about this concern. Although of course, upgrading SQL Express to a later version is always an option too.

[pufferdude]

Thanks for the insights and heads up!

[Regnor]

@Anton: Well you're right and I also don't think that this will be a problem, but it also doesn't hurt to go with a supported SQL version

[Gostev]

This thinking hurt me before too many times... which is why personally I much prefer "don't touch what's working" approach sticking to proven stuff until I'm forced out. Saves me so much time lately!

[mikeely]

Our existing VBR instance is on ancient hardware and is scheduled for a refresh. We're currently running Server 2016 which, aside from eye-wateringly long Windows Update times (I have to schedule 45 minutes downtime and have run over) it seems to run well enough.

Our repo is NFS and will remain so for the time being, so ReFS is not a consideration.

Those of you who have run VBR on multiple different Server versions, what was your experience? Is there any reason I should prefer/avoid any version? Thanks!

[mikeely]

I was hoping to hear about different Server versions and what experiences people were having with them, not whether or not I should pick one particular version

[Gostev]

For a brand new server, there have to be some very good reasons not to pick this particular version, and instead go with some outdated one (that will go out of support sooner) so this thread is perfect to hear and collect these reasons

Also, I'm not sure what version-specific "experiences" you're hoping to hear about in the first place. OS does not matter for Veeam Backup & Replication: the software works equally well on any supported Windows version and we have literally a few hundreds of thousands of active installations on both Server 2016 and Server 2019. So if there were any version-specific issues or peculiarities, there would be a very long topic about them already.

Server 2022 is quite new on the other hand, and this is the perfect topic to share any knowledge why it should not be used solely for deploying Veeam in 2022, if anyone has such opinion.

[dali@iae.nl]

No problems VBR 11a on Win2022 except for 1 issue: "Failed to list S3 buckets" on AWS. Same config and credentials on earlier Windows versions no problemen (I also just put this issue in the Object Storage area).

[EDIT] It seems to be a timing issue as found in the C:\ProgramData\Veeam\Backup\Satellites\VBR\VBR_Administrator\Agent.PublicCloud.Satellite.log file --> "3 error: The difference between the request time and the current time is too large."
Corrected the exact time and now I can list the S3 buckets. Case closed.

So still no reason no to upgrade tot Windows 2022 - it is a super fast OS.

[Regnor]

This thinking hurt me before too many times... which is why personally I much prefer "don't touch what's working" approach sticking to proven stuff until I'm forced out. Saves me so much time lately!

At least I can't say anything bad about SQL, which seems to be the only Microsoft product which doesn't fail regularly. But I'm not a database administrator and I'm sure you can tell different stories

@Mike: It's 2019 vs 2022 I would say. So far I haven't heard anything bad about 2022. I have it in my lab and will do a production deployment in a few weeks.

[Gostev]

No, I agree... SQL Server has been rock solid. It's a great piece of software.

[oscaru]

Hi,

We deal with multiple Veeam B&R installations, both 2016 and 2019 are solid, after applying updates that fix all those ReFS bugs, almost everybody knows about.

Last month we deployed a new backup server with Windows Server 2022, and Veeam B&R 11a, and it is working great, so I recommended like a solid choice to deploy new servers, and you get the most current os, with the last ReFS version, and more years of support.

We did follow this excellent blog post on installing an SQL Server 2019 instance before installing B&R 11a:

https://community.veeam.com/blogs-and-p ... t-one-1499

[ddayton]

If you are running on a pre Windows 2019 server just the Windows update factor alone is enough to make the upgrade worth it. 2016 is horrible for applying updates, so thank you MS for 2019. I upgraded from 2012 to 2016 early in 2020 and just before installing VBR 11.0 I upgraded to 2019. So far I have never had any stability or problems due to the Windows Server with Veeam B&R or Veeam ONE.

[mikeely]

Sounds like 2022 is probably the right path. Somewhat related, right now our VBR/Veeam One installation sits on SQL Server 2017 rather than the built-in Express database. How's the upgrade path look there? I know that's going to have an impact on our hardware refresh since SQL Server licenses by core rather than socket - the XEON Gold 6250 sure looks appealing. Is there even a reason to consider downgrading to SQL Express?

[Mildur]

Is there even a reason to consider downgrading to SQL Express?

I don‘t recommend to use sql express for veeam One.
Depending on your backup/infrastructure size, veeam one will collect a large amount of data. SQL express has a DB limitation of 10GB.

For the vbr migration, install a newer sql server and use veeam configuration backup to migrate the configuration backup.

[mikeely]

Thanks Mildur, was trying to remember why we'd switched up in the first place.

[Mildur]

Your welcome.
I added a link on how to migrate the vbr servers database.
And here the link for the Veeam One database

[sleander]

I migrated my vbr server to a 2022 machine about 2 weeks ago, so far so good.

[RubinCompServ]

This thinking hurt me before too many times... which is why personally I much prefer "don't touch what's working" approach sticking to proven stuff until I'm forced out. Saves me so much time lately!

Agreed! Which is why my VBR is running on Windows 2008 R2...


(just kidding!)

[Gostev]

And going Server 2022 for new backup servers today will allow to "not touch what's working" for longer than with any other OS version

[m.novelli]

I have many Windows Server 2022 and they works fine with Veeam

Marco

[edison5000]

After reading a bunch of stuff, including the B&R Best Practice guide, a variety of posts here and elsewhere, I still don't have a good idea of whether to implement ReFS on our brand new, still being built B&R server.

We're installing 11a.
We're not using Storage Spaces.
Our primary area for repository is a big RAID 60'ed 16 disk array, which I'll format accordingly.
We are only backing up hardware endpoints (desktops & notebooks). No vm's, no servers (probably), no duplication.

I've read about bugs and problems around 2018 and prior, with Veeam specifically, but don't seem much post-2018.

ReFS - use it? Don't use it? Why not?

Thanks veeamers. : )

[edison5000]

Hola. Not sure why I got moved to this particular thread.

I found this kb from 2018

https://www.veeam.com/kb2792
ReFS Known Issues, Considerations, and Limitations

but says it was updated 4/2022. Then I found a Reddit thread from 2021 where Gostev said:
https://www.reddit.com/r/Veeam/comments ... s_for_nas/

"Bottom line, I would go ReFS/XFS for primary backup repository regardless of storage hardware. First, as you admitted, it's been working well for you for years now, so why change now? Second, there are too many potential events that may result in a loss of your primary repository, and you need to be ready for this regardless by making copies of your backups to another repository or cloud."

This is what I was looking for. But - it is not directly mentioned this succinctly in the B&R forum FAQ. The B&R forum FAQ references blog post below as to ReFS advantages, but it's from 2016:

Advanced ReFS integration coming in Veeam Availability Suite 9.5
https://www.veeam.com/blog/advanced-ref ... suite.html

So I get that Veeam is advocating "use ReFS, it's best", but not exactly sure that "right now in 2022, the potential pitfalls are the same as they were in 2018."

[m.novelli]

After reading a bunch of stuff, including the B&R Best Practice guide, a variety of posts here and elsewhere, I still don't have a good idea of whether to implement ReFS on our brand new, still being built B&R server.

We're installing 11a.
We're not using Storage Spaces.
Our primary area for repository is a big RAID 60'ed 16 disk array, which I'll format accordingly.
We are only backing up hardware endpoints (desktops & notebooks). No vm's, no servers (probably), no duplication.

I've read about bugs and problems around 2018 and prior, with Veeam specifically, but don't seem much post-2018.

ReFS - use it? Don't use it? Why not?

Thanks veeamers. : )

I have dozens of Veeam Backup Server running on Dell hardware, Windows Server 2019 or Windows Server 2022 , no issues with ReFS 64K

Volume size range from 21 to 72 TB RAID6

Marco

[edison5000]

Thanks Marco. : )

[chris.childerhose]

The performance on Win2022 ReFS is much better than in previous versions. We have tested many servers with large TB of data. No issues moving to this version.

[Ctek]

After reading a bunch of stuff, including the B&R Best Practice guide, a variety of posts here and elsewhere, I still don't have a good idea of whether to implement ReFS on our brand new, still being built B&R server.

We're installing 11a.
We're not using Storage Spaces.
Our primary area for repository is a big RAID 60'ed 16 disk array, which I'll format accordingly.
We are only backing up hardware endpoints (desktops & notebooks). No vm's, no servers (probably), no duplication.

I've read about bugs and problems around 2018 and prior, with Veeam specifically, but don't seem much post-2018.

ReFS - use it? Don't use it? Why not?

Thanks veeamers. : )

From a Cloud service provider perspective, with petabytes of Veeam storage, and with lots or experience and hits/misses in ReFS storage, here's my dirty TLDR:

-Follow the 1TB Storage:1GB of RAM ratio golden rule. It may be debated with the latest updates, but it still is IMO a golden rule. Anyways, with newer Xeon SP balanced RAM configurations with 6 and 8 channels, you gotta have a lot of RAM to be efficient right off the bat...

-NEVER EVER REBOOT a repository server unless you are 100% absolutely sure it is NOT doing ReFS storage tasks in the background. Most of the ReFS storage operations are not easily detectable unless you take a close look with task manager/ process manager for the background processes. Others can chime in on that comment, but i've seen guys get RAW'ed 100+ TB partitions while patching repository servers with /force, no recovery possible even with MS Support to this day.

-If you have set registry keys for ReFS optimization in the past on said repository server, , unless very knowledgeable on the matter, remove them and monitor your servers if there is issues. Sometimes this can complexity troubleshooting, especially after patching.

-If you use hardware RAID, hopefully RAID60 for larger arrays, and you need to rebuild a disk, I highly suggest to throttle down the backups.

I know it's not on the topic, but we are also using Linux XFS repositories in our premises and so far, we are quite happy.

Cheers from Canada.

[edison5000]

Thanks Ctek. So, nope, I have no experience with ReFS. I am brand new to Veeam and backup server stuff in general. I hope not to touch the registry at all? And yet...

Our rig for this brand new mission, from scratch, is -

brand new 24core EPYC AMD server
64gb ram
480gb BOSS 2 disk mirror - Win server 2022 + Veeam parts (B&R, BEM, maybe Veeam one)
480gb 2 disk mirror on perc controller - SQL instance / databases
16 14tb disks in a locally attached RAID 60 24 bay array (that gives us around 150TB) - repository (this would be the ReFS chunk)
10gb network card

will backup around 800 desktops & notebooks, no other devices, all on premise, no cloud access or need to connect to other locations.

You're saying we should have 150GB of ram? In any case, we won't have a separate repository server; everything Veeam is happening stacked on this one box. I didn't exactly architect this, but it's what we got.

My other concern right now (though i'm deviating from my own thread) is whether we have enough space on the primary disk to handle all logs & catalog. I know after my last dev attempt, installing every dang Veeam product (B&R up to VSPC) - all Veeam things plus Server 2022 only took up around 80gb, so 400 was still left. (Granted that was before generating any backups). Since SQL is on its own disk, it seems like it might be ok if I tell Veeam to write over old logs after a certain point. I found this veeam kb on editing registry to mod the log sizes (not sure why this can't be in B&R gui?), but am wary of doing that.

https://www.veeam.com/kb1825