The openvswitch run in between is broken even without running against proposed.
Actually it is the external network configuration which hands me no IP via dhcp anymore.
That leads to issues doing apt-updates from the guests once connected directly to the external network via OVS or OVS-DPDK.
With some manual pondering I got the OVS-DPDK based run with vhostuserclient connections (the one we most care for) done:
3.2.1 (07:52:15): test guest-dpdk-vhost-user-client-multiq for OVSDPDK-VUC => Pass
^^ sounds like 50 minutes later, but that took 3 days to get working :-/
1. Just as we do with migration-reference/0 in autopkgtest I have verified that those sub-cases fail with proposed just as much.
2. We know from the many runs so far that it also affects not only non-proposed but also pure OVS runs (without DPDK in the mix).
3. Furthermore it seems mostly flaky, failing on different stages of the test each time.
Gladly Jammy and Focal (more important) were fine and fully complete before this happened.
I'd consider kinetic verified to the extend it can be verified at the moment.
If you object, then just wait until kinetic is EOL and update just all the others (I think that would be silly, but fulfill the requirements without spending too much time while being blocked on this).
The openvswitch run in between is broken even without running against proposed.
Actually it is the external network configuration which hands me no IP via dhcp anymore.
That leads to issues doing apt-updates from the guests once connected directly to the external network via OVS or OVS-DPDK.
I've now run all other tests, they are good:
1.0.0 (07:00:18): phys (BM) tests
1.1.0 (07:00:18): initialize environment
1.1.1 (07:01:35): testpmd => Pass
1.1.2 (07:02:24): check testpmd output => Pass
2.0.0 (07:02:24): prep virtual test environment
With some manual pondering I got the OVS-DPDK based run with vhostuserclient connections (the one we most care for) done: vhost-user- client- multiq for OVSDPDK-VUC => Pass
3.2.1 (07:52:15): test guest-dpdk-
^^ sounds like 50 minutes later, but that took 3 days to get working :-/
1. Just as we do with migration- reference/ 0 in autopkgtest I have verified that those sub-cases fail with proposed just as much.
2. We know from the many runs so far that it also affects not only non-proposed but also pure OVS runs (without DPDK in the mix).
3. Furthermore it seems mostly flaky, failing on different stages of the test each time.
Gladly Jammy and Focal (more important) were fine and fully complete before this happened.
I'd consider kinetic verified to the extend it can be verified at the moment.
If you object, then just wait until kinetic is EOL and update just all the others (I think that would be silly, but fulfill the requirements without spending too much time while being blocked on this).