VMware/Xen to Proxmox: display emulation issues during boot

[RaphNTX]

Hello,

We have VMware and Xen (Linux guest) VM's we would like to migrate to Proxmox but we encounter issues during the boot sequence of these VM's, once converted. The disks have been converted with xenmigrate, qemu-img to get a qcow2 disk file.

When we boot a converted VM, we typically see this on the VM console (noVNC):
1. seaBIOS prompt and welcome to Grub prompt
2. screen replaced by Grub blue boot selection menu (kernel version, rescue mode, etc...)
4. after selecting an entry or waiting the selection timeout, screen is replaced by "booting ...", "loading kernel..."
5. then, it switches back to the screen in step 1 (we see seaBIOS prompt and welcome to Grub prompt)

I think that the guest OS switch to a resolution which is not supported by Proxmox because if I interrupt Grub to edit the boot parameters and remove the graphical console options in order to have pure text-based boot sequence, then the VM boots fine.

The problem is that we have hundred's of VM's to convert and editing the Grub settings one-by-one is just unrealistic.

The issue is present with both Debian and CentOS VM's.

We also tried the different display emulations options in Proxmox (standard VGA, vmware, ...) but they all fail.

Has anyone encountered this issue?

Best Regards,
Raph

 

[LnxBil]

We also tried the different display emulations options in Proxmox (standard VGA, vmware, ...) but they all fail.

Has anyone encountered this issue?

Only problems that could be solved by switching to vmware or editing grub cmdline. Why Xen fails is very strange, normally XEN uses QEMU for their hardware emulation layer.

Can't you change all entries on all servers currently running and migrate them afterwards? With so many servers you will probably have some SSH-magic-run-on-all-nodes-scripts, don't you?

 

[RaphNTX]

Only problems that could be solved by switching to vmware or editing grub cmdline. Why Xen fails is very strange, normally XEN uses QEMU for their hardware emulation layer.

Can't you change all entries on all servers currently running and migrate them afterwards? With so many servers you will probably have some SSH-magic-run-on-all-nodes-scripts, don't you?

We tried display mode default, standard VGA and vmware already, without success.

Unfortunately, we don't have global SSH magic to update all VM's automatically as they are managed by different teams in an heterogeneous environment.

 

[dcsapak]

you can try with cirrus, but you have to set that on the cli with

qm set ID -vga cirrus

but since this is very old and there were some serious vulnerabilities, we changed the default to 'std' and removed it as an option from the gui

 

[RaphNTX]

you can try with cirrus, but you have to set that on the cli with

qm set ID -vga cirrus

but since this is very old and there were some serious vulnerabilities, we changed the default to 'std' and removed it as an option from the gui

Unfortunately, we already tried that one too, without success