QA » openQA Project » openQA Tests

PR: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/14681

Can confirm the same situation happening on all ppc64le HA migration tests:
https://openqa.suse.de/tests/8494904#step/boot_to_desktop/8

I am wondering if it is related to the 'virtio-rng-pci' mentioned in : https://bugzilla.suse.com/show_bug.cgi?id=1198205

Lumir.

lpalovsky wrote:

Can confirm the same situation happening on all ppc64le HA migration tests:
https://openqa.suse.de/tests/8494904#step/boot_to_desktop/8

I am wondering if it is related to the 'virtio-rng-pci' mentioned in : https://bugzilla.suse.com/show_bug.cgi?id=1198205

Lumir.

Use my PR, it can work: https://openqa.suse.de/tests/8503560#step/boot_to_desktop/6

@lpalovsky, please help to review PR: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/14681

tinawang123 wrote:

@lpalovsky, please help to review PR: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/14681

For me it seems like setting QEMU_VIRTIO_RNG=0 does the job.
https://openqa.suse.de/tests/8503840#

It probably fixes https://bugzilla.suse.com/show_bug.cgi?id=1198205 too.
Not sure what was the purpose of this new rng device.

lpalovsky wrote:

It probably fixes https://bugzilla.suse.com/show_bug.cgi?id=1198205 too.
Not sure what was the purpose of this new rng device.

Hardware machines have an rng, so we would like to be as close as possible in testing.
Also on aarch64 there are KASLR related errors (KASLR doesn't work without rng there) without RNG.

dheidler wrote:

lpalovsky wrote:

It probably fixes https://bugzilla.suse.com/show_bug.cgi?id=1198205 too.
Not sure what was the purpose of this new rng device.

Hardware machines have an rng, so we would like to be as close as possible in testing.
Also on aarch64 there are KASLR related errors (KASLR doesn't work without rng there) without RNG.

Yes, I agree. It's good to have it covered. The issue here is a new device being added to the configuration. For most of the scenarios this is ok but migration tests are using pre-installed qcows which are then migrated to the target version. This new device then causes naming shift and results in various failures:

1. power machines trying to boot incorrect device - This can be fixed with test code in PR @tinawang123 proposed. The code was using a static value so it's good to improve it anyway.However this will still lead to the second issue.
2. during migration, yast starts to complain about boot devices and this affects all HA power and x86 based scenarios. https://openqa.suse.de/tests/8495019#step/start_install/2

This is harder to fix. You either need to re-create all source qcow images with new rng device included, or adding a function to the testcode which would proceed in the migration.
Disabling rng device in yaml only for affected scenarios and architectures seems like a good compromise.