We are trying to use the live migration capability to move VMs off of an old ubuntu precise (12.04) KVM host.
We are using the --copy-storage-all option to move block devices during the migration, and it seems that the old version of QEMU (v1.0) on precise copies the block device into RAM as part of the migration. This typically results in RAM running out on the host machine causing OOM-killer to shoot the VM being migrated, and the migration failing. Is there a way to stop this happening? It appears not to happen in newer QEMU versions, so I wonder if it is a bug with the old QEMU version, but I have been unable to find any mention of one. Sadly, the entire point of live migrating VMs off is to allow us to upgrade the host OS to a newer version. -- Tom Denley
