BEWARE: possible RAW ReFS volumes after installing January updates

[Gostev]

Thank you Gostev. Appreciate you pointing me to the right place. I read through the 2 pages here. Our repositories are running 2019. It seems like I would be ok to move forward with the patches at this point?

Unless you're using USB drives as backup media, you should not have any issues in the first place. With rare exceptions caused by OS seeing fixed drives as external drives.

[stfconsulting]

Thank you very much for taking the time to explain this to me. Your efforts save us all a lot of time / pain. Keep it up!

[pesos]

So, Microsoft has a KB article up about this, which states that ReFS is not supported on external drives:

https://support.microsoft.com/en-us/top ... 9b0ee86c65

I hope that they intend to fix this. Despite it being "unsupported," (something I didn't know, and I'm not sure how I could have been expected to know since it's not like Windows warned you if you tried to format an external drive with ReFS), it has worked perfectly fine for years. I have a bunch of USB external drives we rotate through as backup repositories and to move all of the data, reformat to NTFS, and copy it back would be extremely time consuming.

Curious if anyone has learned anything further regarding this aspect. We are in a similar situation with some sites with USB-based refs repositories.

[agrob]

Unless you're using USB drives as backup media, you should not have any issues in the first place. With rare exceptions caused by OS seeing fixed drives as external drives.

I can confirm that i had no issue with installing the patches last week on a HPE Apollo 4510 with local ReFS Volume (Raid 60)

[APangF1]

To summarize, the January 2022 cumulative update problem affects both removeable drives and fixed drives (direct attached, iscsi) with ReFS 2 or lower. (check ReFS version using fsutil fsinfo refsinfo <drive letter> )

Anyone know if the February 2022 cumulative update has the same problem ?

[Gostev]

No, only removable drives.

[FrancWest]

Sorry, but @ApangF1 is correct, see my previous posts in this thread. On vSphere you might run into the issue too when using a virtual harddisk which still has ReFS version 1.2 for some reason.

[Gostev]

His statement is not correct: the issue does NOT apply to fixed drives, but only to drives that the OS sees as removable. As far as I understood from ReFS developers, there's a storage class check in the ReFS driver code during the mount process to avoid mounting removable media. Sure, a particular virtualization platform may present virtual "fixed" drives as removable to the guest OS, causing the issue to trigger on such removable drives. But this does not mean the issue itself applies to fixed drives.

[FrancWest]

Ok, ok , but most of us think of removable drives as external (USB) drives, not as 'fixed' drives being presented as removable drive to the guest OS by vSphere. That's why I made my comment to clarify things that the issue also might occur with vSphere vmdk's. I don't know if this issue also occurs with other hypervisors.

[pesos]

Sooooooooo... curious what others who have been running usb-based ReFS repositories for some time are doing about this.

We paused updates for a while hoping this decision would be reversed or a workaround made available for scenarios where such drives are basically fixed and could be designated as "trusted" - but nothing yet.

In order to get patches going again, we've gone ahead and switched a number of small community edition sites to ntfs - generally just a single vm is being backed up at these locations and there's no data, just a print/imaging server that it would be a pain to lose.

But we have a handful of other sites with more important usb-based ReFS repos... ones where fast cloning was factored into the performance and space considerations so shifting to ntfs isn't an easy move.


EDIT: also curious how an ReFS-formatted VHDX sitting atop an NTFS-formatted USB drive would behave...

[soncscy]

> EDIT: also curious how an ReFS-formatted VHDX sitting atop an NTFS-formatted USB drive would behave...

I would avoid this It's too many layers of abstraction and if something _does_ go wrong, now you're troubleshooting on the hardware layer, ntfs, refs, and the vhdx systems.

If your goal is to get block cloning savings on the cheap, I'd look into investing into a moderately priced old box and put Ubuntu Server on it and use XFS. The system requirements should be pretty low so you can even likely grab an old decommissioned workstation and stand it in for this purpose. It's an extra box, sure, but it's stable, highly supported, very simple to set up. I've not had an environment where we tried external drives with any Linux box, but I can't find any major complaint threads which a cursory look online (I advise research though on your own)

ReFS is great _when it works_, but for my taste just too many exceptions keep appearing about it over the years. I wouldn't say "don't" use it, but at this point for Veeam, if you're at a point when you're going to remake your repository anyways, I would push for an XFS based solution in virtually all situations that allow it.

[APangF1]

We need more info from Microsoft on the exact conditions (which we would then be able to check) that turn ReFS into RAW volumes. We'll hold on windows updates for all ReFS repositories, since we are planning a transition to immutable with XFS. Thanks

[stephc_msft]

See the big discussion under
https://community.veeam.com/discussion- ... ndex4.html
(Not sure why the Veeam Communities and Forum are different things ...)

[Gostev]

It's "R&D Forum". Veeam Communities were added to host all non-R&D related stuff... this forum was becoming very busy even with "on-topic" content and because of its popularity people started bringing all sorts of random stuff here which had nothing to deal with R&D.

[matej.blaskovic]

Franc, do you also use a USB hard drive for ReFS repository?

Very good question > and actually all scsi discks on VM ware are indeed removable/ hotplug
- I will be testing a scenario where we will disable hotplug for discs and will patch - I will come back later if that helped

[matej.blaskovic]

Yes, it did help
Insert a new row with the name devices.hotplug and a value of false to the VM options > Advanced > Edit configuration