CentOS 7.3 VM file level recovery issues

[Salm]

Hello there !

First of all, here is the environment :

• Veeam 9.5 Update 1 (9.5.0.711)
  Hyper-V 2016 Datacenter fully updated (as of today)
  CentOS 7.3 VM fully updated as well (7.3.1611)
  VM is Gen 2, configuration 8.0, and with secure boot enabled
  Filesystem is XFS, hosted in a LVM (default CentOS install settings)
  Guest-processing was not enabled - Hyper-V guest quiescence was used (there's only files on this VM anyways)

Here's the problem :

When I use the file level recovery, I can recover a file without any issue by using the "Copy To..." option from the restore, to restore a file to my Veeam server or my desktop running the console



However, if I use the "Restore" menu, with either option ("Overwrite" or "Keep"), I get this error "No HyperV integration services are installed in the VM"


This VM is a CentOS 7.3 guest OS, which has the Linux Integration Services built in. According to this article, I do not need to install anything else (and the VM is running perfectly, with all drivers recognized).

Am I missing something ?

Thanks !

Edit : the current kernel is 3.10.0-514.6.1.e17.x86_64. I tried updating to 4.9.9-1.el17.elrepo, but it doesn't have LIS included yet (so no network, and no integration services).

[Mike Resseler]

Hi Salm,

Is this machine reachable from the network by your server where you perform the file level recovery?

[Salm]

Hi,

Yes, all the machines (VM, FLR appliance, Veeam Server, Veeam console (Desktop)) are on the same subnet.

However, the firewall is activated on the VM, and I noticed from this KB that I may need TCP ports 2500-5000 open for it to work.

Port 22 is obviously open, but do I require this range as well ?

Another thing, when I try to get the VM IP using the integration services through Hyper-V, I get nothing


It may be related ?

[Mike Resseler]

Hmm,

Degraded, ProtocolVersion normally means that the IC's (whether it is Linux or Windows) are not up-to-date and different compared to what the host expects. But because you are running 2016, it should be OK since it now comes with Updates (for Windows) or built-in LIS as you have in your specific distro.

Unfortunately I am not sure how this can be resolved. You might want to get MSFT support on this one

Mike

[Salm]

Yeah, I wasn't worried about the "Degraded" part. What is weird is that it cannot get the IP address from the integration services (which, by the way, breaks application aware processing as Veeam cannot get the guest's IP address).

I guess it's just another problem with Hyper-V 2016 ? I will have to try with a 2012 R2 host to confirm.

I'll push the issue with support if I get a client with the problem (this is in my homelab, so not really critical : I can live with the "Copy to..." restore method).

Thanks !

[Salm]

Well, I just tried on a 2012 R2 Hyper-V Host, and the issue is identical (no IP from integration services, error when trying to restore directly to guest).

I guess the problem comes from the implementation of LIS in CentOS 7.3.

[Mike Resseler]

So a support call to MSFT it is I guess

Thanks for coming back with the results of your test

Mike

[rkovhaev]

Linux HV integration components consist of two parts: user-space HV daemons and kernel-space HV modules.

kernel-space HV modules are part of mainstream Linux kernel - if you are running fairly new Linux kernel they will be loaded automatically.
user-space components you will need to install manually (package is usually called hyperv-daemons).

here is how to verify:

kernel-space modules:
[root@rklinux ~]# lsmod |grep hv
hv_utils 24418 2
hv_netvsc 40045 0
hv_storvsc 18032 2
hv_vmbus 397185 6 hyperv_keyboard,hv_netvsc,hid_hyperv,hv_utils,hyperv_fb,hv_storvsc

user-space services:
[root@rklinux ~]# ps aux |grep hyper
root 620 0.0 0.1 37960 1460 ? Ss 22:36 0:00 /usr/sbin/hypervkvpd -n
root 621 0.0 0.0 4292 564 ? Ss 22:36 0:00 /usr/sbin/hypervvssd -n

[Salm]

Well, I think that it's working now, the user-space components were not installed.

Well, the file restore failed, but because TCP port 2500 is not open.
The restore was attempted (and it can now read the VM's IP).

I find it weird that I didn't find anything on this.
Seems you specifically have to google for hyperv-daemons to get information on it...

In any case, thank you !