Hello,
We're in a uncommon situation where we're moving CTs between two datacenter with a limited link (latency + BW)
All nodes are connected in the same ceph cluster, with different "locals" pools.
We noticed that moving CTs storage (from source pool to the new one, doing the storage move after the VM has been migrated to the new hypervisor) with a lot of small files is slow, probably dues to the limited link, but not using the full bandwidth at all.
Latency changes between cases, with a bigger impact with a bigger latency.
This doesn't seem to occur when moving VMs, but it's seem the migration is not done the same way: rsync for CTs (with both pools mounted on the current hypervisor), and raw block copy (?) for VMs ?
Is there a way we could do that better ? The GUI is quite limited when moving disks :/
We know running ceph thought the limited link is not a good idea, this is only for the migration
Thanks a lot,
We're in a uncommon situation where we're moving CTs between two datacenter with a limited link (latency + BW)
All nodes are connected in the same ceph cluster, with different "locals" pools.
We noticed that moving CTs storage (from source pool to the new one, doing the storage move after the VM has been migrated to the new hypervisor) with a lot of small files is slow, probably dues to the limited link, but not using the full bandwidth at all.
Latency changes between cases, with a bigger impact with a bigger latency.
This doesn't seem to occur when moving VMs, but it's seem the migration is not done the same way: rsync for CTs (with both pools mounted on the current hypervisor), and raw block copy (?) for VMs ?
Is there a way we could do that better ? The GUI is quite limited when moving disks :/
We know running ceph thought the limited link is not a good idea, this is only for the migration
Thanks a lot,