Open josecastillolema opened 1 year ago
This might have unforeseen consequences of time-skew, possibly expired certificates, and who knows what else... plus not all hypervisors support this. We already have/had some issues on macOS when the machine went in suspend/hibernate and the cluster would not continue properly. though, at the moment I 'paused' my own Microshift VM using cockpit and see what happens when resumed. Also, not sure why but the guest agent does not start in the Microshift image. This I also noticed yesterday when debugging an issue around the startup phase.
Also, not sure why but the guest agent does not start in the Microshift image. This I also noticed yesterday when debugging an issue around the startup phase.
We configure qemu-guest-agent to use vsock for communication, but we don't add a vsock device to our VM at least on linux iirc, maybe this is what you saw.
maybe this is what you saw.
Does this prevent the guest-agent from functioning properly, as the service on Linux-hosted VMs fails.
maybe this is what you saw.
Does this prevent the guest-agent from functioning properly, as the service on Linux-hosted VMs fails.
If you start the agent with --listen vsock://...
but the VM does not support vsock devices, yes the guest agent will fail to start.
Is there any other use of the guest-agent? if not, perhaps we have to remove it, or fix the start definition, as at the moment it shows 'degraded' as state from systemd.
The only user of qemu-guest-agent is macos timesync. With recent changes in snc
, we could add it only to the macos bundle. On the other hand, should be a simple change to always add vsock to our VMs.
similar tools do not provide a pause either; prefer to use stop and start (and the optimization could be to speed up the 'restart' process).
It would be great to implement support for pausing the CRC virtual machine. In development laptops, where CRC consumes a significant amount of resources/batteries, this would allow for an easy (and MUCH quicker than stop/start) pause/resume mechanism to interact with the local OCP environment. A couple of failed attempts to pause the VM form the
virsh
perspective: