"Daniel P. Berrange" <berra...@redhat.com> wrote: > On Thu, Jan 11, 2018 at 01:23:05PM +0000, Dr. David Alan Gilbert wrote: >> * Daniel P. Berrange (berra...@redhat.com) wrote: >> > On Thu, Jan 11, 2018 at 01:46:38PM +0100, Max Reitz wrote: >> > > On 2018-01-08 14:52, Eric Blake wrote: >> > > > On 01/07/2018 06:23 AM, Richard Palethorpe wrote: >> > > >> Add QAPI wrapper functions for the existing snapshot functionality. >> > > >> These >> > > >> functions behave the same way as the HMP savevm, loadvm and delvm >> > > >> commands. This will allow applications, such as OpenQA, to >> > > >> programmatically >> > > >> revert the VM to a previous state with no dependence on HMP or >> > > >> qemu-img. >> > > > >> > > > That's already possible; libvirt uses QMP's human-monitor-command to >> > > > access these HMP commands programmatically. >> > > > >> > > > We've had discussions in the past about what it would take to have >> > > > specific QMP commands for these operations; the biggest problem is that >> > > > these commands promote the use of internal snapshots, and there are >> > > > enough performance and other issues with internal snapshots that we are >> > > > not yet ready to commit to a long-term interface for making their use >> > > > easier. At this point, our recommendation is to prefer external >> > > > snapshots. >> > > >> > > We already have QMP commands for internal snapshots, though. Isn't the >> > > biggest issue that savevm takes too much time to be a synchronous QMP >> > > command? >> > >> > Ultimately savevm/loadvm are using much of the migration code internally, >> > but are not exposed as URI schemes. Could we perhaps take advantage of >> > the internal common layer and define a migration URI scheme >> > >> > snapshot:<name> >> > >> > where '<name>' is the name of the internal snapshot in the qcow2 file. >> >> I had wondered about that; I'd just thought of doing the migration >> saving to a block device rather than the rest of the snapshot >> activity around it; >> but I guess that's possible. > > One possible gotcha is whether the current savevm/loadvm QEMUFile impl > actually does non-blocking I/O properly. eg same reason why we don't > support a plain file:<path> protocol - POSIX I/O on plain files doesn't > honour O_NONBLOCK. The block layer does AIO though, so we might be OK, > depending on which block layer APIs the QEMUFile impl uses. I've not > looked at the code recently though.
The blocking part is less important (for the write side), because we have a thread there. For loading .... it would be great to get one migration thread also. >> > Then you could just use the regular migrate QMP commands for loading >> > and saving snapshots. Might need a little extra work on the incoming >> > side, since we need to be able to load snapshots, despite QEMU not >> > being started with '-incoming defer', but might still be doable ? >> > This would theoretically give us progress monitoring, cancellation, >> > etc for free. >> >> What actually stops this working other than the sanity check in >> migrate_incoming ? > > No idea really - not looked closely at the code implications. It would be a plus for migration code, right now there are _two_ implementations, and savevm/loadvm one gets less love. And we will check "much more" the way to load migration in a non-pristine qemu, so .... Later, Juan.