QA » openQA Project » openQA Tests

could be related to my recent changes in https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/4585 :(

afaics, the problem was introduced with my change regarding more robust reconnect -> PR

I've added the execution of assert_shutdown_and_restore_system to S390-KVM:

-    if (check_var('VIRSH_VMM_FAMILY', 'xen')) {
+    if (check_var('VIRSH_VMM_FAMILY', 'xen') || get_var('S390_ZKVM')) {
         assert_shutdown_and_restore_system($action, $shutdown_timeout);
     }

Here I've put in the logic of the former redefine_svirt_domain, which is modifying the xml file

 491         if (check_var('ARCH', 's390x') or get_var('NETBOOT')) {
 492             $svirt->change_domain_element(os => initrd  => undef);
 493             $svirt->change_domain_element(os => kernel  => undef);
 494             $svirt->change_domain_element(os => cmdline => undef);
 495             $svirt->change_domain_element(on_reboot => undef);
 496             $svirt->define_and_start;
 497         }

On initial definition in the xml files, the on_reboot flag is set to destroy, because we need to remodify the xml file - after this we set it to undef

Since we are booting an existing image in the toolchain testsuite, this parameter is by default set to reboot.

The assert_shutdown function which is called afterwards, tries to grep for shut off of the machine, which obviously fails, because the machine was rebooting just normally.

So my guess is, that this issue will occur on all tests which are rebooting again during the testsuite execution. As we don't have that many, and those we have failed earlier, this wasn't (yet) spotted there.

I try to fix this issue by adding another logical statement to that call... or I find a better solution to handle those reboots