So, this morning I took a bunch of time to poke around on a jenkins which had seen this failure. I couldn't see an obvious smoking gun, apart from this log message which implies that the qemu monitor might have crashed:
2013-11-19 21:52:20.053+0000: 9833: error : qemuMonitorIORead:513 : Unable to read from monitor: Connection reset by peer
Another thing I've tried is I've run a bunch of stable/havana reviews through check, and none of them have failed so far with 10 runs. That makes me think this isn't something that's changed in libvirt or qemu, although perhaps I am wrong.
I'm now at the point of reading the diff between stable/havana and master for the libvirt driver. The most obvious thing I see there that's changed is that we've started using lifecycle events. I wonder if that exposed a bug in libvirt?
So, this morning I took a bunch of time to poke around on a jenkins which had seen this failure. I couldn't see an obvious smoking gun, apart from this log message which implies that the qemu monitor might have crashed:
2013-11-19 21:52:20.053+0000: 9833: error : qemuMonitorIORe ad:513 : Unable to read from
monitor: Connection reset by peer
Another thing I've tried is I've run a bunch of stable/havana reviews through check, and none of them have failed so far with 10 runs. That makes me think this isn't something that's changed in libvirt or qemu, although perhaps I am wrong.
I'm now at the point of reading the diff between stable/havana and master for the libvirt driver. The most obvious thing I see there that's changed is that we've started using lifecycle events. I wonder if that exposed a bug in libvirt?