When migrating a VM (regular one, not live) OpenNebula does the following:
- Pauses the VM by saving its current state
- Copies that VM with its saved state to a different host
- Resume the VM
The problem with this is that when the VM is resumed, the clock on the VM continues from where it left of. Having ntpd
/chronyd
running on a guest doesn’t help, because they have some max time diff value at which they don’t adjust the time. This is especially a problem for big VMs which can take an hour (or even more) to migrate.
How do you overcome time sync problems when migrating VMs?