On 03/13/2012 05:00 PM, Ante Karamatic wrote:

Hi all

I've noticed libvirt's qemu hook doesn't make a difference between crashed and stopped VM. While I do understand that crashed VM is essentially a stopped VM, I'd be interested in providing (and working on) a patch that would differentiate these two cases.

What I'm interested in is if there's a reason, unknown to me, why this wasn't implemented in the first place. And of course, does it make sense to differentiate those two.

Thank you!

There are reasons defined for a stopped VM: typedef enum { VIR_DOMAIN_SHUTOFF_UNKNOWN = 0, /* the reason is unknown */ VIR_DOMAIN_SHUTOFF_SHUTDOWN = 1, /* normal shutdown */ VIR_DOMAIN_SHUTOFF_DESTROYED = 2, /* forced poweroff */ VIR_DOMAIN_SHUTOFF_CRASHED = 3, /* domain crashed */ VIR_DOMAIN_SHUTOFF_MIGRATED = 4, /* migrated to another host */ VIR_DOMAIN_SHUTOFF_SAVED = 5, /* saved to a file */ VIR_DOMAIN_SHUTOFF_FAILED = 6, /* domain failed to start */ VIR_DOMAIN_SHUTOFF_FROM_SNAPSHOT = 7, /* restored from a snapshot which was * taken while domain was shutoff */ #ifdef VIR_ENUM_SENTINELS VIR_DOMAIN_SHUTOFF_LAST #endif } virDomainShutoffReason; A rough thought is you can take use of the parameter "@extra" to pass the a string to represent the shutoff reason. <snip> * @op: the operation on the id e.g. VIR_HOOK_QEMU_OP_START * @sub_op: a sub_operation, currently unused * @extra: optional string information <snip> Osier