Instant recovery of VM with large RAM + reservation fails with "insufficient disk space"

[mdiver]

It seems VBR checks for the availability of vPower NFS cache space equalling a full swap file for the recovered VMs full RAM allocation.
This is even done with a memory reservation at 100% of the RAM which would not produce the swap at all.

I've just realized IR jobs of 1.5TB RAM HANA VMs fail for that reason as we don't have the vPower NFS cache at >1.5TB.
Is that by design? Veeam could check for RAM reservation in advance.

Thanks,
Mike

[HannesK]

Hello,
VMware creates a .vswp file on the datastore with the amount of RAM. So yes, that's by design from VMware.

When I try to start a VM on a datastore that has less free disk space than I assign RAM to that VM, then I get exactly the message you see from VMware side.

Veeam could check for RAM reservation in advance.

Can you explain? It's a VMware requirement to have enough free space for the .vswp file

Best regards,
Hannes

[mdiver]

It's not a requirement with active RAM reservation.
Then only an empty .vswp is generated.
This is especially crucial for HANA workloads, as it is SAPs best practice and also normally you don't have the amount of RAM also as extra space on your production storage (1.5TB+).

With IR Veeam seems to check for the full RAM amount - no matter if a reservation is active or not.
This prevents an IR of large HANA workloads as usually you don't have several TB of NFS cache. Especially because it won't get used.

Thanks,
Mike

[HannesK]

Hello,
hmm okay... just to be sure that I understand the question correctly: you are talking about this checkbox? "reserve all guest memory (all locked)"



by checking that box, I see that the .vswp is empty. It is also empty for IVMR (checkbox is set). So I guess, that I still do not have the right configuration to reproduce the issue.

Best regards,
Hannes

[mdiver]

Hi Hannes.
Yes, that's the checkbox. Correct, the .vswp is empty. As the IVMR inherits the setting, it's also empty. This is expected.
It's just that VBR checks for enough space on the vPower NFS cache for the vswp. This is done no matter if it will be generated or not. This is my problem. It wont be generated for a HANA workload. But I still cannot fire up the VM from IR because of this check.
Only if I use a redirect for the cache to productive storage, the check seems to be circumvented.
Thanks,
Mike

[HannesK]

Hi Mike
thanks, now I got it and reproduced it.

That's a valid request for improvement. I will check with my colleagues

Best regards,
Hannes