The Friday 21 Feb 2014 à 15:09:42 (-0700), Eric Blake wrote : > On 02/21/2014 02:21 PM, Benoît Canet wrote: > > From: Benoît Canet <ben...@irqsave.net> > > > > This patchset enables the core of the quorum mechanism. > > The num_children reads are compared to get the majority version and if this > > version exists more than threshold times the guest won't see the error at > > all. > > > +++ b/docs/qmp/qmp-events.txt > > @@ -500,3 +500,39 @@ Example: > > > > Note: If action is "reset", "shutdown", or "pause" the WATCHDOG event is > > followed respectively by the RESET, SHUTDOWN, or STOP events. > > + > > +QUORUM_FAILURE > > +-------------- > > This should have been inserted in sorted order to match the rest of the > file, rather than at the end. Can be rearranged in a followup patch, > now that this series is queued. > > > + > > +Emitted by the Quorum block driver if it fails to establish a quorum. > > + > > +Data: > > + > > +- "reference": device name if defined else node name. > > +- "sector-num": Number of the first sector of the failed read operation. > > I'm assuming sector-num is relative to the guest's view of the data? Yes,
> > > + > > +QUORUM_REPORT_BAD > > +----------------- > > + > > +Emitted to report a corruption of a Quorum file. > > + > > +Data: > > + > > +- "ret": The IO return code. > > What values is this likely to contain? Is it a finite set, in which > case it would be nice to have a QAPI enum that describes the set of > return codes, rather than a raw number? It's anything that the block stack could return as an error. > > I'd like to make sure we are happy with the wire format of this QMP > event before it gets baked into 2.0. Ok > > -- > Eric Blake eblake redhat com +1-919-301-3266 > Libvirt virtualization library http://libvirt.org >