On Thu, Dec 03, 2020 at 07:11:17AM -0500, Michael S. Tsirkin wrote: > On Thu, Dec 03, 2020 at 11:43:41AM +0000, Dr. David Alan Gilbert wrote: > > Another way to solve this would be to remove the unplugging from the > > migration layer and leave it as a problem for the management layer to do > > the unplug. > > Daniel described the problem with modular management tools which expect > pausing or slowing down guest to cause migration to converge. > > Point is, it actually *will* make it converge but only if you > pause it after unplug. > > As it is, these tools fundamentally can not handle failover > requiring guest cooperation. Moving code between layers won't help. > Introducing failure modes as this patch does won't help either > especially since Daniel wrote there are countless tools like this. > We just break them all but have no resources to fix them, > this does not help at all. > > We can just leave the situation as is. > > Or if we do want to be nice to these tools, how about we > unpause the guest until unplug, then pause it again? > This actually addresses the problem instead of > shifting the blame, does it not?
This is a very bad idea because it changes the execution status of the guest behind the apps/admins back, and that cannot be assumed to be a safe thing todo. Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|