I did some further autopkgtest triage for the latest `ovn` runs triggered by
openvswitch/3.3.9-0ubuntu0.24.04.1 (noble). None of the failures are regressions
caused by openvswitch 3.3.9.

Runs analysed:
- s390x: 
https://autopkgtest.ubuntu.com/packages/ovn/noble/s390x/20260727_101837_9e3ab@
- amd64: 
https://autopkgtest.ubuntu.com/packages/ovn/noble/amd64/20260727_100645_37d42@

regress-stack functional (tempest) verification, run against openvswitch 3.3.9
from noble-proposed, PASSED with no failures. The relevant output is pasted 
below
(source: https://github.com/canonical/regress-stack/actions/runs/28375371239).

Package under test (noble-proposed job):
  openvswitch-common/noble-proposed  3.3.9-0ubuntu0.24.04.1 amd64 [installed]
  openvswitch-switch/noble-proposed  3.3.9-0ubuntu0.24.04.1 amd64 [installed]
  python3-openvswitch/noble-proposed 3.3.9-0ubuntu0.24.04.1 amd64 [installed]

Result (noble-proposed):
  ======
  Totals
  ======
  Ran: 108 tests in 654.2273 sec.
   - Passed: 98
   - Skipped: 10
   - Expected Fail: 0
   - Unexpected Success: 0
   - Failed: 0

For comparison, the baseline run against openvswitch from noble (release) gives
the identical result (Ran: 108, Passed: 98, Skipped: 10, Failed: 0), confirming
no functional regression from 3.3.9.

Failing autopkgtests and analysis:

"egress qos" (85, 86)
- Fails deterministically on both amd64 and s390x at the same assertion, so it
  is not a testbed glitch. It is NOT caused by openvswitch 3.3.9: it is a
  pre-existing brittleness in the OVN test expected `tc` output that no longer
  matches on noble's kernel/iproute2.

  The test greps for a fixed HTB class string, and the `cburst 3.....b` part
  (a 6-digit value starting with 3) is what fails. `cburst` is computed by the
  kernel from the "unlimited" ceil rate (34359Mbit) and is kernel-dependent.

  Local reproduction on noble (24.04.4, iproute2-6.1.0):

   $ sudo ip link add ovs-testq type veth peer name testq-peer
   $ sudo tc qdisc add dev ovs-testq root handle 1: htb
   $ sudo tc class add dev ovs-testq parent 1: classid 1:1 htb \
      rate 12kbit ceil 34359738360bit burst 375000b
   $ tc class show dev ovs-testq
     class htb 1:1 root prio 0 rate 12Kbit ceil 34359Mbit burst 375000b cburst 
1649b

  The OVN-controlled values are correct (rate 12Kbit, ceil 34359Mbit,
  burst 375000b): only the kernel-computed `cburst` (1649b here) no longer
  matches the hard-coded `cburst 3.....b` pattern. This is an OVN test-assertion
  issue, out of scope for this openvswitch SRU, and will be fixed separately in
  ovn.

Requested action:
- For "egress qos", apply a badtest hint
  (force-badtest ovn/24.03.6-0ubuntu0.24.04.1) so openvswitch can migrate. The
  OVN test failure is pre-existing and unrelated to openvswitch 3.3.9.

Furthermore, here is a passing run with trigger 
netplan.io/1.1.2-8ubuntu1~24.04.2
(previously failing):
https://autopkgtest.ubuntu.com/packages/netplan.io/noble/amd64/20260727_093512_3acfa@

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2147329

Title:
  [SRU] openvswitch 3.3.9 point release

To manage notifications about this bug go to:
https://bugs.launchpad.net/cloud-archive/+bug/2147329/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to