SureBackup & Latency-Sensitive VMs

[micoolpaul]

Morning everyone,

I've got a customer that has fully embraced software defined networking and UC. This means some key components of their network such as firewall, SBC and UC solution are all virtual.

Whilst some VMs such as firewall are a poor candidate for SureBackup, their UC system suits as it is Windows based and has many services that can be queried to confirm they're operating successfully.

The issue I've discovered is that as part of the vendor's best practice for VMware is that they request their latency sensitivity to be set to high and ensure all CPU & RAM is reserved. When the SureBackup task then executes I hit an error that the CPU reservation is below the minimum required for high latency sensitivity.

I've raised a case with Veeam Support but wondered if anyone else had seen this issue previously?

[HannesK]

Hello,
can you please share the case number? Where exactly do you see the CPU reservation error? In SureBackup, or in the application?

If you manually set (script) a CPU reservation on the SureBackup VM, does that solve the problem?

By the way, which UC vendor are you using? When I worked in that business, all vendors did not allow VM based / snapshot based backup. Of cause snapshot restore was also not supported. I'm interested to hear, whether the situation has changed.

Best regards,
Hannes

[micoolpaul]

Evening,

Case number: 04007997

So this customer is using Skype for Business Server & Audiocodes SBCs. VMware's documentation for SfBS recommends CPU reservations which is what they've set alongside the high latency sensitivty flag. Support are currently asking for me to remove the RAM % reduction applied, which I'm going to be testing on tonight's trial run. Whilst Veeam doesn't natively integrate with SfB, as there are SQL instances on the front ends it makes sense that we leverage the application integration for those elements, which is why we want to ensure the SQL server components restore successfully.

I don't believe we'll get as far as executing the script as VMware won't boot the VM due to the CPU reservation being lower than required. Customer is on vSphere 6.7 U3 latest update if you wish to replicate.

Thanks

[HannesK]

Hello,
I can reproduce it. As soon as I set "Latency Sensitivity" to "high" in VMware, I get an error message similar to yours.


In my case it is a memory message, but I assume it's the same issue.

Code: Select all

Error: Invalid memory setting: memory reservation (sched.mem.min) should be equal to memsize

I talked to support, but I don't expect a quick fix for that.

Best regards,
Hannes

[micoolpaul]

Hi Hannes,

Thanks for reproducing! Will keep you updated on any responses from support myself!

Thanks,
Michael

[micoolpaul]

I've had support confirm issue persists in v10 and is with R&D. Thanks Hannes for your help.

[jerseyconnect]

Is there update on this issue?

We just set latency sensitivity on some our VMs this week and we're running into this exact issue with SureBackup. For comparison I tried doing a normal VM restore and did not run into this issue. The manually restored VM has the correct CPU reservation while the SureBackup VM has no CPU reservation configured.

[HannesK]

Hello,
chances are good that the issue (bug ID #85691) will be resolved with the next update (10a)

Best regards,
Hannes

[jerseyconnect]

We updated to 10a yesterday and the issue is still present. I ran a test SureBackup job and the VM is still being created without any CPU reservation configured. The job fails with "Error: Invalid CPU reservation for the latency-sensitive VM".

[HannesK]

Hello,
Hmm, it works fine for me in 10a.

Could you maybe open a case (including logs) and refer to this thread and the bug number?

Thanks,
Hannes

[jerseyconnect]

OK, ticket opened 04439330.

[HannesK]

just to summarize for future readers: there is a different bug that is planned to be fixed with V11

[micoolpaul]

Hi,

I noticed that there was no mention of this bug within the Veeam v11 release notes, can you confirm if it's been fixed but undocumented please?

[Gostev]

Big fixes are never documented as there are thousands of them. Judging on the time of the post (well before RTM) the bug should have been fixed in V11.