Hi,

Disclaimer here: I’ve never used IBM Spectrum Scale. From what I can see however the operating systems it can run on are supported. So I’m essence, Veeam would allow you to create a backup job targeting one or more nodes within this solution.

However from what I can see Veeam have no specific support for this. So this raises the questions around how IBM Spectrum Scale handles being processed. If it was deployed on Windows (as one scenario I saw showed this), Veeam would look to leverage VSS where possible for application consistency, of which I can’t see IBM Spectrum Scale supports this.

So you could either look to IBM to find out what scenarios they support backups being taken of their nodes (you can run scripts pre and post processing which may be sufficient to “stun” a node whilst being processed) or you could carry  out a crash consistent backup and see how IBM Spectrum Scale would handle bringing this node back on line.

In summary, as much as I hate the phrase “it depends”.

Veeam can install on the OS and as a result create backups, but really we don’t care about taking backups we care about the recoverability of data, we’ll need more information to assist further!

As far as Spectrum Scale is backed up as a exported filesystem - like in the scenario when Spectrum Scale is backed up with the “normal” Spectrum Protect Backup/Archive Client - everything works ok.

But there is no special Spectrum Scale Client like mmbackup. For example there is no support for the Spectrum Scale journaling functions with Veeam.

Mr. @MicoolPaul  and Mr. @JMeixner 

Thank you so much for your reply on my query.

As much I have tested veeam b&r software for backup and recovery of linux systems using IBM spectrum scale (GPFS), the backup process of volume (using either mount point or device) is successful but it's recovery fails and makes the volume unreadable, even dismount the volume.

I have tested the following scenarios:-

Scenario 1: Backup and recovery of data (using mount point while taking backup) which is replicated using IBM spectrum scale

- Mount 2 LUN from 2 different SAN (1 TB LUN from SAN-1 and 1 TB LUN from SAN-2) on a Linux server. Devices name of LUN are /dev/dm-0 and /dev/dm-1, respectively.

- Create a single mount point /U01 for these 2 LUN and did data replication between these LUN using IBM spectrum scale (GPFS v.5.1.1.0). The /U01 now shows 2 TB size (actual 1 TB data which is replicated between the 2 underline LUN)

- Now when I take backup using volume mount point (/U01) the backup job completes by creating a backup of 2TB size.

_ My questions is:

     • Why veeam fails in recognizing the data? As the actual data is 1TB, which is replicated between 2 SAN, so the backup size must also be of 1TB instead of 2TB. Also, this is impossible to restore this 2 TB backup on 1 TB actual SAN.

Scenario 2: Backup and recovery of data (using device while taking backup) which is replicated using IBM spectrum scale

- In order to overcome the problems as mentioned in Scenario 1, here in this scenario, backup is taken using device (i.e., /dev/dm-0), as device /dev/dm-1 has the replicated data of /dev/dm-0.

- The backup job completes by creating of backup of 1 TB.

- Now when I restore the data either on the same volume, i.e., /dev/dm-0 or another volume, e.g., /dev/dm-2, the restoration completes but makes the volume unreadable. Means when try to run the 'll' or 'ls' command, an error of "input/output error" occurs. Also, when I reboot the system the volume is dismounted. 

_ My questions is:

    • Why the error occurs while reading/displying the restored volume? Also, why the restored volume is dismounted?

Scenario 3: Backup and recovery of data (file system is GPFS v. 5.1.1.0)

- In this scenario, data is not replicated but the file system GPFS. 

- Volume Backup job runs successfully (both using mount point and device while creating backup job).

- Restored process runs successfully (both on same volume and another volume), but when reading/displying the data in the restored volume, the same errors of above Scenario 2 (i.e., input/output error while running 'll' or 'ls' command, and also, when I reboot the system the volume is dismounted).

Finding:

From the above scenarios, I think, veeam  does not recognize GPFS.

​​​​Solicitation: 

Experts suggestions are requested on my tested scenarios and finding, please.

Could there be an issue with the multipathing of this device?

Veeam can read from one device. But if it tries to write to this device, then there seems to be confusing with the replication.

How do you backup the data? As a file backup or as a volume backup?

I think, only file backup will be possible...

Mr. @JMeixner ,

• Multipaths are OK, tried both Multipahts of Linux as well as NetApp multipath drivers.
• Sir, it fails even we try to restore backup to volume on which replication is not applied.
• For backup, we use volume backup.

Did you try file level backup and restore?

Mr. @JMeixner , I haven't tried file level backup and restore. Will try it and update you accordingly.

Hello Asad, I’m working on a similar use case, maybe you tried to backup and restore at file level ?  can you share your experience ?  Regards

Hi,

I have tried to backup on file level. Works without problems. For restore it seems that you have to use the host on which the file share resides as location for the helper host.

I prepare a detailled post on this at the moment.

Thank Jochen very intresting and I can't wait  for read your post as soon as 

it will take a week or two, I am rather busy at the moment, have to renew the water pipes in my house… but it will come here in community.