Am 2022-02-09 um 13:02 schrieb Simon Lees:
On 2/9/22 21:39, Axel Braun wrote:
Hi Lubos,
Am Mittwoch, 9. Februar 2022, 11:09:00 CET schrieb Lubos Kocman: ....
## openSUSE Tumbleweed
* Kernel 5.16.4 in Tumbleweed, 5.16.5 will be in 0207 (QA running), 5.16.7 in staging (with simple drm enabled)
While there was a discussion about performance improvments in OBS in todays RelEng-Meeting, I seem to remember that we talked about this for openQA some time ago as well.
Looking at the current 20220207 snapshot, the only pending QA jobs are for flavours that are out of maintenance since...years partly: https://paste.opensuse.org/31618724
In order speed testing up I would like to start the discussion about outdated test cases...where should this be discussed?
Unfortunately these probably aren't out of date yet traditionally for SLE atleast upgrading by skipping a major version has been supported. Which means in practice its likely that SUSE will want to support upgrades from SLE-12-SP5 (or whatever we get to) to SLE-16 (or whatever its called).
In practice at the moment upgrading a system from 42.X to Tumbleweed is the closest approximation we can do in openQA for that use case.
There are multiple benefits for these upgrade jobs, among others 1. Simply show that it actually still works to upgrade from old Leap to Tumbleweed :) Maybe some users can still use that as a reference to upgrade their completely outdated and unsupported older openSUSE to a current openSUSE Tumbleweed ;) 2. As a "stress-test" doing a lot of package replacement and shuffling around of files
So I guess a counter question is, what is the need for speeding up testing? are we at the point where sometimes openQA tests run for greater then 24hrs so new snapshots are released invalidating the testing on the previous one, or are we running low on resources for other tests to run?
At least for x86_64 and aarch64 I don't think we commonly have that problem on openqa.opensuse.org. Also https://metrics.opensuse.org/d/osrt_openqa/osrt-openqa?orgId=1&from=now-30d&to=now shows that the queue of pending jobs often enough reaches zero meaning there is enough capacity. We even use the free capacity to run automatic investigation jobs and provide the instance for developing tests by individual developers so again, there is enough capacity :)