Can VMs migrate live with process running?

[jjc27017]

Hi all,
I am running proxmox as Virtual Environment 5.1-35 with ceph as storage. And it runs very well as fit as my company. But I want to know the thing is do vms online migration can not migrate with the process running at that time? I have been set a group of HA and connect with some vms. When the physical server down then the virtual hosts will migrate, but they run on other physical node with nothing like restarted.

I know I can migrate after I hang the vms. And after migration then recover the virtual hosts can keep the process running. But the physical server will occur error and shutdown in some time I don't know. Please tell me how can HA migrate with hosts include their running processes. Or the function has not yet be come true?

 

[Belokan]

As far as I known, the only way to to what you want is to permanently "sync" the memory image of the VMs on each nodes of the HA group. I guess it is technically possible but not yet implemented. And to do so, you'll need a network infrastructure correctly sized to achieve it without a great loss in global performance.

So right now, if you can manually migrate a VM before the host crashes (I/O errors, oom killer in action, whatever, ...) it is "live" and running processes will keep running. But if the host crashes when the VM is actually running on it, it'll be restarted, I mean like if it crashed itself including filesystems recovery and so on, on the next available node ...

 

[jjc27017]

As far as I known, the only way to to what you want is to permanently "sync" the memory image of the VMs on each nodes of the HA group. I guess it is technically possible but not yet implemented. And to do so, you'll need a network infrastructure correctly sized to achieve it without a great loss in global performance.

So right now, if you can manually migrate a VM before the host crashes (I/O errors, oom killer in action, whatever, ...) it is "live" and running processes will keep running. But if the host crashes when the VM is actually running on it, it'll be restarted, I mean like if it crashed itself including filesystems recovery and so on, on the next available node ...

Thanks for answering, so I think the vm migration is the same as lxc migration, it eventually restart the virtual hosts.

 

[LnxBil]

Thanks for answering, so I think the vm migration is the same as lxc migration, it eventually restart the virtual hosts.

That's wrong. You can online migrate a KVM host without downtime without a problem. This is not possible with LXC.

Migration (intentionally moving) works without a problem, fault-tolerant behavior does not (e.g. node fails and everything runs smoothly on another node without any disruption). In the later case the VM will be started, as the normal LXC does.

 

[jjc27017]

That's wrong. You can online migrate a KVM host without downtime without a problem. This is not possible with LXC.

Migration (intentionally moving) works without a problem, fault-tolerant behavior does not (e.g. node fails and everything runs smoothly on another node without any disruption). In the later case the VM will be started, as the normal LXC does.

Is there any ways that migration can also migrate ‘running processes’?

 

[Belokan]

Is there any ways that migration can also migrate ‘running processes’?

That's what migration does. Migration requires source and target nodes up and running.

 

[LnxBil]

Is there any ways that migration can also migrate ‘running processes’?

Please define what you mean by processes? For me, this implies LXC and this is currently not possible, as I wrote earlier.