[ovirt-devel] VM rebooted during OST test_hotplug_cpu

Hi all, I am trying to verify fixes for ovirt-engine-rename, specifically for OVN. Engine top patch is [1], OST patch [2]. Ran the manual job on these [3]. In previous patches, OST failed in earlier tests. Now, it passed these tests, so I hope that my patches are enough for what I am trying to do. However, [3] did fail later, during test_hotplug_cpu - it set the number of CPUs, then tried to connect to the VM, and timed out. The logs imply that right after it changed the number of CPUs, the VM was rebooted, apparently by libvirtd. Relevant log snippets: vdsm [4]: 2020-06-29 10:21:10,889-0400 DEBUG (jsonrpc/1) [virt.vm] (vmId='7474280d-4501-4355-9425-63898757682b') Setting number of cpus to : 2 (vm:3089) 2020-06-29 10:21:10,952-0400 INFO (jsonrpc/1) [api.virt] FINISH setNumberOfCpus return={'status': {'code': 0, 'message': 'Done'}, 'vmList': {}} from=::ffff:192.168.201.4,54576, flow_id=7f9503ed, vmId=7474280d-4501-4355-9425-63898757682b (api:54) 2020-06-29 10:21:11,111-0400 DEBUG (periodic/0) [virt.sampling.VMBulkstatsMonitor] sampled timestamp 2925.602824355 elapsed 0.160 acquired True domains all (sampling:451) 2020-06-29 10:21:11,430-0400 DEBUG (jsonrpc/1) [jsonrpc.JsonRpcServer] Return 'VM.setNumberOfCpus' in bridge with {} (__init__:356) 2020-06-29 10:21:11,432-0400 INFO (jsonrpc/1) [jsonrpc.JsonRpcServer] RPC call VM.setNumberOfCpus succeeded in 0.56 seconds (__init__:312) 2020-06-29 10:21:12,228-0400 INFO (libvirt/events) [virt.vm] (vmId='7474280d-4501-4355-9425-63898757682b') reboot event (vm:1033) qemu [5]: 2020-06-29T14:21:12.260303Z qemu-kvm: terminating on signal 15 from pid 42224 (<unknown process>) 2020-06-29 14:21:12.462+0000: shutting down, reason=destroyed libvirtd [6] itself does not log anything relevant AFAICT, but at least it shows that the above unknown process is itself: 2020-06-29 14:18:16.212+0000: 42224: error : qemuMonitorIO:620 : internal error: End of file from qemu monitor (Note that above line is from 3 minutes before the reboot, and the only place in the log with '42224'. No other log there has 42224, other than these and audit.log). Any idea? Is this a bug in libvirt? vdsm? I'd at least expect something in the log for such a severe step. [1] https://gerrit.ovirt.org/109961 [2] https://gerrit.ovirt.org/109734 [3] https://jenkins.ovirt.org/view/oVirt%20system%20tests/job/ovirt-system-te... [4] https://jenkins.ovirt.org/view/oVirt%20system%20tests/job/ovirt-system-te... [5] https://jenkins.ovirt.org/view/oVirt%20system%20tests/job/ovirt-system-te... [6] https://jenkins.ovirt.org/view/oVirt%20system%20tests/job/ovirt-system-te... -- Didi