At Thu, 13 May 2010 05:16:35 +0900, MORITA Kazutaka wrote: > > On 2010/05/12 23:28, Avi Kivity wrote: > > On 05/12/2010 01:46 PM, MORITA Kazutaka wrote: > >> This patch calls the close handler of the block driver before the qemu > >> process exits. > >> > >> This is necessary because the sheepdog block driver releases the lock > >> of VM images in the close handler. > >> > >> > > > > How do you handle abnormal termination? > > > > In the case, we need to release the lock manually, unfortunately. > Sheepdog admin tool has a command to do that. > More precisely, if qemu fails down with its host machine, we detect the qemu failure and release the lock. It is because Sheepdog currently assumes that all the qemu processes are in the sheepdog cluster, and remember that where they are running. When machine failures happen, sheepdog release the lock of the VMs on the machines (We uses corosync to check which machines are alive or not). If the qemu process exits abnormally and its host machine is alive, the sheepdog daemon on the host needs to detect the qemu failure. However, the feature is not implemented yet. We think of checking a socket connection between qemu and sheepdog daemon to detect the failure. Currently, we need to release the lock manually from the admin tool in this case. Thanks, Kazutaka -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
