Created attachment 1417246 [details]
engine, libvirt and vdsm logs
Description of problem: HA VM with lease, that was paused with i/o error, is restarted to the active host, but doesn't get OS
Version-Release number of selected component (if applicable):vdsm, engine, libvirt logs.
Steps to Reproduce:
1.Run HA VM (bootable iscsi disk) with lease on iscsi_0 SD.
2.block the iscsi on the host where VM is started. Wait >80sec
Actual results: VM is started on another active host, but no OS. No IP
Expected results: VM is started on another active host, got OS.
Additional info: logs attached: engine.log (vm name='golden_env_mixed_virtio_2_1' - see at the end of the log). libvirt and vdsm from the host to where the VM was restarted and didn't get the IP
we need guest logs, can you please add those? Or check using graphical console what's going on in the guest after restart.
Does it boot correctly when you run the same VM (normally, from Down) on that host?
Created attachment 1418583 [details]
the guest log is attached
in the console, there is booting stuck.
the same VM starts on the same host from Down randomly successfully, which could be related to another bug . but when VM is restarted as result of i/o error pause, it always has the OS problem
could be related to https://bugzilla.redhat.com/show_bug.cgi?id=1562750
since that bug 1562750 reproduces about 50% can you tell if this particular issue really happens all the time or roughly 50%?
Maybe you can try with different disk configuration? (or even just a diskless VM) and reproduce that problem independently to HA
Tested on rhv-release-4.2.4-5-001.noarch, vdsm-4.20.31-1.el7ev.x86_64, ibvirt-3.9.0-14.el7_5.5.x86_64
The problem is not reproduced. VM is restarted and get IP.