Dedicated snapshot storage feature in 13.x

[brf]

In the "What's New" PDF for VBR 13, Linux agent section, one of the new features mentioned is the following:

Dedicated snapshot storage — The agent can now be configured to use the dedicated block device as the snapshot storage. This enables storing snapshot data directly on a raw (unformatted) disk, removing performance overhead of a file system. And by using the dedicated disk, you can be sure that snapshot storage will never impact available production storage capacity.

I can't find this mentioned anywhere in the agent documentation, and there also don't seem to be any references to this feature in the new agent's default veeam.ini. So, how does one actually use this feature?

[rovshan.pashayev]

Hi,

You can use the existing configuration line (under snapshot section) in the veeam.ini file as an example:

Code: Select all

[snapshot]
...
# Location folder for snapshot data, only for 'stretch' and 'common' snapshot
# location = /dev/sdX1

Replace /dev/sdX1 with the appropriate device for your environment.
Previously, this configuration line was only used for specifying directories, but now it supports block devices for dedicated snapshot storage.

[brf]

Yeah, I'm using it for directories now. I suspected that maybe I just needed to put a device name in that parameter, but the explanation text above it in the default veeam.ini says this: "Location folder for snapshot data, only for 'stretch' and 'common' snapshot"

I'm thinking the "Location folder" wording there needs to be updated, as that's what discouraged me from trying it.

[rovshan.pashayev]

Hi

Thanks for the feedback! We’ll review this and improve the labelling.

[brf]

I'm trying to use this now but getting this error: "Cannot use the block device '/dev/sdc' as a difference storage because it needs to be backed up."

Previously, we had /dev/sdc1 mounted as an empty filesystem on /snapshots, and then set "location = /snapshots". Now, I unmounted /dev/sdc1 and then wrote zeros onto /dev/sdc (to ensure that the OS wouldn't see it as a partitioned disk or a filesystem), then set "location = /dev/sdc" in veeam.ini.

For the agent backup job, I have it set to back up the "Entire computer", which I would prefer to keep (to make sure that filesystems that get added in the future don't get missed).

How would I fix this? Is there something I can set in veeam.ini to ignore /dev/sdc as a backup target?

Also just tried creating a partition /dev/sdc1 (unformatted) and use that, just in case the behavior was any different...but same error.

[brf]

I did some digging and found that there's an "ignoreDevices" parameter in the [backup] section of /etc/veeam/veeam.ini, which isn't part of the default veeam.ini. So I added a line for that, containing /dev/sdc. That got me a different error: "The block device [/dev/sdc] cannot be used as a difference storage because it's busy." Don't know what that would mean.

[brf]

Another interesting note here: after removing my old snapshots filesystem that was on /dev/sdc1, the next time I rebooted, Linux decided to swap my /dev/sdb and /dev/sdc devices. So, if Veeam had actually went ahead and used the device name that I specified as a snapshot (fortunately it didn't), it would have blown away critical data on the device that is now /dev/sdc, but used to be /dev/sdb at the time I set up the configuration file.

It seems extremely unsafe to specify a snapshot device in this way. I would think we need some way to specify the UUID of a partition, or maybe put a Veeam snapshot signature onto a particular device/partition, or something else along those lines.

edit: guess we could create an LVM PV on the device in question, then create an LV and use it by name as the device? Like /dev/VG1/SNAPSHOT-LV or whatever. Seems like it would be the only safe way to use this feature?

[rovshan.pashayev]

Hi,

it would have blown away critical data

Veeam Agent will not overwrite a device that contains existing data or non‑Veeam signatures—that’s why you’re seeing the “The block device [/dev/sdc] cannot be used as a difference storage because it's busy.” error.

Please ensure you’re pointing to the correct device using a location and ignoreDevices entries.

If the issue persists, please open a support case and share the case number here for reference.

[brf]

Yeah, it still claims it's "busy" even when location and ignoreDevices are both set correctly.

I guess without knowing what in particular it's looking for to determine whether it contains "existing data", it's impossible to know what we need to do to make it usable. Zero'ing out the first 128MB of the device apparently isn't good enough, because I did that.

Even if it won't overwrite the contents of something that "contains existing data" (you have to hope it gets that right), it's really still not safe to specify devices by simple /dev/sda style names because your backups are now broken if your hardware/OS changes the device names on you. LVM LV's seem like the only safe path unless Veeam gives us something else like UUIDs.

I'll open a support ticket about both these topics later today. Thanks

[brf]

By the way, if ignoreDevices is necessary to make this work, then another issue is having to set ignoreDevices by device path to begin with.

For example:

Let's say I want to use /dev/sdc as my snapshot disk. So I set ignoreDevices=/dev/sdc.

Then I reboot, and because of hardware changes or OS upgrades, maybe a disk is offline, or even just Linux's whim, it decided to change some device ordering. Now /dev/sdc contains critical data. Now /dev/sdc isn't being backed up anymore, and I don't even realize.

All of that is too dangerous to use.

[brf]

Hmm, another safe option could be to use the symlinks that are in the /dev/disk/by-uuid or /dev/disk/by-partuuid directories.

[brf]

Support case is #07948181