Hi
I have installed PVE 8.4.1 with vGPU (Nvidia A16) on two Servers in a Cluster, The tested Driver Versions of Nvidia are 18.0 and 18.1.
Now if I have to reboot or shutdown these Host Servers i'm getting the message "watchdog0: watchdog did not stop! \n systemd-shutdown[1]: Failed to finalize DM devices, ignoring."
I know that the Timeout of the systemd watchdog is 10min, but the Server will also not completely stop or restart after 30min, so I have to force shutdown the Server every time. The strange thing is, if I'm removing the GPU (Nvidia A16) from the Server, but not uninstalling the GPU Drivers, the Host will shutdown or reboot normally.
We also have other Clusters with vGPU with PVE 8.3.2/8.3.3 and Nvidia Driver 17.5, with these Versions we never had these issue.
Is these a Problem on the Linux side or do I have to raise a Nvidia Ticket?
Thanks.
Btw: Big thank you to Proxmox for Implementing the Live Migration Feature of vGPUs, it works very well so far, only the Migration with Bulk Migration will not work.
I have installed PVE 8.4.1 with vGPU (Nvidia A16) on two Servers in a Cluster, The tested Driver Versions of Nvidia are 18.0 and 18.1.
Now if I have to reboot or shutdown these Host Servers i'm getting the message "watchdog0: watchdog did not stop! \n systemd-shutdown[1]: Failed to finalize DM devices, ignoring."
I know that the Timeout of the systemd watchdog is 10min, but the Server will also not completely stop or restart after 30min, so I have to force shutdown the Server every time. The strange thing is, if I'm removing the GPU (Nvidia A16) from the Server, but not uninstalling the GPU Drivers, the Host will shutdown or reboot normally.
We also have other Clusters with vGPU with PVE 8.3.2/8.3.3 and Nvidia Driver 17.5, with these Versions we never had these issue.
Is these a Problem on the Linux side or do I have to raise a Nvidia Ticket?
Thanks.
Btw: Big thank you to Proxmox for Implementing the Live Migration Feature of vGPUs, it works very well so far, only the Migration with Bulk Migration will not work.
Last edited: