Expanding on viewtopic.php?f=2&t=104316&start=0, as I did some testing around with Instant Recovery onto PVE.
After the mentioned thing in regard of sudo, I could actually kinda do an Instant Recovery of a virtual machine into "local" once I have allowed "Disk Images" to be stored there.
Once the VM was actually running, I noticed, that there was no NFS-Datastore anywhere to be seen, which is different to vSphere, ehere there is actually a NFS-Datastore mounted. On investigating on that system itself, i noticed the reason, why there is actually access to "local", or more precisely a file based storage, required:
Instant Recovery creates a QCOW2 file, which references a VMDK file residing in a NFS mount bellow /var/nfs as "backing file". This is IMHO kinda dirty, but anyhow, It works (shrug). I have done worse
So, the next step is migrating to production, and here I have some real issues, which IMHO need to be addressed:
The first option is simply do this via Veeam itself. But the issue here is, that the VM can only be migrated to the storage, that had been selected during the "instant recovery" dialog at the beginning. If this is not big enough, the whole thing fails.
Which bring us to the second option, which in line with vSphere, is migrating the VM via the Proxmox-UI itself. In vSphere this is no issue, as Veeam detects, that the VM had been migrated from the NFS mount, and simply finishes the IR-jobs successfully. But with Proxmox, this detection does not work. Even after several Minutes the job is still visible as "running". So I stopped the IR-job, but in that case, _Veeam deleted the already migrated VM within Proxmox_, even though I enabled "Delete Source on completion" enabled for the Migration done with Proxmox.
_This_ is IMHO completely against the "Principle of least Astonishment". Having a big enough file based storage available is not always feasible, especially for big systems like file or exchange servers. So Veeam should be at least able to detect, that the VM had already been migrated, and is not residing on the temporary NFS-Storage any more.
IMHO this makes IR at the moment borderline unusable for PVE workloads.
As a traditional restore, i.e. "Full VM recovery", is already possible to RAW storages like CEPH, it should be possible ask the restore operator for the destination storage for "Migration to production".
Cheers
Jörg