[BETA] [HPE] iSCSI disks are being enumerated in guests

[dcolpitts]

Previously, in ESXi where we had a standalone host, we would deploy a Ubuntu VM, configure iSCSI on it, and present a LUN from a NAS or something as iSCSI to it. We would then turn this VM into our hardened repository, and write the Veeam backups to the iSCSI LUN. With Veeam and ESXi when we added the repository VM to the back up job, the total size of the VM was only what was actually in the VMDKs on the host. I just added our harden repository VM to a new backup job, and the size of the VM being shown is including the iSCSI volume (total 54.5TB), not just the sum of the virtual disk that makes it up (20GB).

When the backup runs, it does only grab the virtual disk. However it's a bit misleading and concerning when you see the size of your backup job is 54.5TB instead of the 10.1GB it actually should be.

dcc

[cody.ault]

Are you configuring iSCSI at the host level and presenting it as a RAW device or inside the guest OS?

[dcolpitts]

It's a mounted iSCSI LUN inside the VM, not the VME host. So basically, take the NAS, create an iSCSI volume on it, and present that volume to the VM. It would be no different than a Nimble or Alletra MP iSCSI volume. You create the volume on the array, and present the LUN to the VM's iSCSI initiator name, format the LUN as xfs from inside the VM once it is discovered and mount it via fstab after installing open-iscsi and multipath-tools.

I can confirm the backup only backups the virtual disk that is defined in the VM. But like I said, it appears VBR is enumerating the VM for all it's storage with the VME plugin. VBR doesn't do this under ESXi.

dcc

[cody.ault]

I'm guessing this might be related to a guest agent reporting size back to qemu/libvirt but I'll try to find out

[EvgenyBaev]

Hello Dean,

Thank you for your valuable feedback.

I was able to reproduce the same scenario in our lab, and I agree - the behavior can indeed be confusing. The root cause lies in how the VM size is calculated: we were retrieving it from the "stats" block in the HPE response, specifically the "maxStorage" field. However, this field includes the size of volumes mounted via network storage (I checked with Alletra and NFS share), which are not displayed in the HPE user interface.

This issue is not limited to the Backup Job wizard; it also affects the Infrastructure tab.

To address this, we have identified an alternative approach for calculating the provisioned VM size. In the upcoming version 1.0, we will aggregate the "maxStorage" values for each volume in the HPE response - excluding CDROM volumes - to more accurately reflect the VM's storage allocation.

Thank you once again for highlighting this issue. Your input is greatly appreciated and helps us improve the product.