QA » openQA Project » openQA Tests

Yeah, the 3 second timeout just was not enough in this case... We can make it 5, or 10 seconds but we can't use check_screen at that point unfortunately.

but we should be able to use an equivalent of wait serial? Would something as in commit efb46071 (lord/fix/zkvm_svirt)
Author: Oliver Kurz okurz@suse.de
Date: Fri Jul 28 13:12:54 2017 +0200

Explicitly confirm grub boot selection in svirt console to save some seconds

Verification run: http://lord.arch/tests/7099

work?

In this case we would need serial console output from the intermediary VM to which we connect, if I understand it correctly, which is possible in this case, but in case of more VMs running concurrently, we might get the magic string belonging to another VM, I am afraid. Retriggering the job or even moving the VM host closer to the worker (i.e. from Prague to Nue) seems more sensible to me...

hi @michalnowak, as a first step I don't mind if you propose a solution that is just blindly waiting longer to make it more robust with a hint to this ticket that this should be improved in the long run.

One thing we could possibly do is to test in select_console('svirt') (when activating it) for match for "Password:" needle (https://openqa.suse.de/tests/1206281#step/bootloader_hyperv/2) and enter VIRSH_PASSWORD from test code.

Was wrong in my analysis above.

Looks to me like the sort of problems fixed by https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/4311:

[2018-02-22T14:13:35.0973 CET] [debug] Command executed: powershell -Command Start-VM openQA-SUT-2
[2018-02-22T14:13:40.0238 CET] [debug] Command's stderr:
Start-VM : The operation cannot be performed while the object is in use.

But an actual verification that hyperv_cmd_with_retry works on scale of build is certainly desirable :).