Hello all,
we are required to migrate all (~100) VMs from an "old" PVE cluster to a new one (both on dedicated servers in Hetzner).
- All VM disks (hard drive /cloud init etc) are on a shared storage (ZFS over iSCSI) that both clusters have mounted successfully.
- A separate PBS instance is available, that has all VMs backed up and is mounted on both clusters.
- Network bridges have the same names.
- VM IPs are failover IPs handled by vSwitches external to PVE
What would the optimum method to efficiently move the VMs across be? To our understanding the options available are:
- Restore from PBS, but this is at the time of the last backup
- Live migrate, but is experimental and to our understanding will copy the drive even though it's already available on the "other side"
- Manually recreate the VM configuration, and select the appropriate disk from storage (too much work)
- Looking into copying VM configuration from `/etc/pve/qemu-server` over, stopping on old and starting on new.
Last option seems best however we do not know the implications of this, hence the question here. Is it sufficient? Are there hidden issues with that approach? We are only interested in migrating the actual VMs, all other configuration (storages/fw/certs/etc) has been created by hand and has been tested on new cluster.
we are required to migrate all (~100) VMs from an "old" PVE cluster to a new one (both on dedicated servers in Hetzner).
- All VM disks (hard drive /cloud init etc) are on a shared storage (ZFS over iSCSI) that both clusters have mounted successfully.
- A separate PBS instance is available, that has all VMs backed up and is mounted on both clusters.
- Network bridges have the same names.
- VM IPs are failover IPs handled by vSwitches external to PVE
What would the optimum method to efficiently move the VMs across be? To our understanding the options available are:
- Restore from PBS, but this is at the time of the last backup
- Live migrate, but is experimental and to our understanding will copy the drive even though it's already available on the "other side"
- Manually recreate the VM configuration, and select the appropriate disk from storage (too much work)
- Looking into copying VM configuration from `/etc/pve/qemu-server` over, stopping on old and starting on new.
Last option seems best however we do not know the implications of this, hence the question here. Is it sufficient? Are there hidden issues with that approach? We are only interested in migrating the actual VMs, all other configuration (storages/fw/certs/etc) has been created by hand and has been tested on new cluster.
Last edited: