neutron-metadata-agent does not sometimes provide instance-id

[Lukas Piwowarski]

Neutron-metadata-agent does not *sometimes* reply to instance-id request by CirrOS VM in OpenStack environment created by Devstack. This causes a failure of upstream jobs (e.g. tempest-full-multinode-py3 [1]) as some tests are not able to successfully finish creation of a VM due to missing response from http://169.254.169.254/meta-data/instance-id (see attached log file).

[1] https://zuul.opendev.org/t/openstack/build/94f0cf28c8fd495d96099f888329dbb2/logs

[Rodolfo Alonso (rodolfo-alonso-hernandez)]

From the logs you have provided, I would first check why the route addition failed:
  WARN: failed: route add -net "0.0.0.0/0" gw "10.1.0.1"

If you don't have the correct routes, you can't access to the metadata service.

[yatin (yatinkarel)]

Seems this is valid issue but in OVN version in ubuntu jammy.

From logs i see OVN versions are below in a jammy job which is impacted by issue mentioned in https://bugzilla.redhat.com/show_bug.cgi?id=2076604:-
ii ovn-central 22.03.0-0ubuntu1 amd64 OVN central components
ii ovn-common 22.03.0-0ubuntu1 amd64 OVN common components
ii ovn-controller-vtep 22.03.0-0ubuntu1 amd64 OVN vtep controller
ii ovn-host 22.03.0-0ubuntu1 amd64 OVN host components

Assuming this issue is seen only with ubuntu-jammy then it's high chance issue is same as mentioned in the above bz as focal had very old ovn which doesn't had this issue. The fix for that issue in 22.03 branch available with 22.03.1+ tags so if we can get it updated in jammy that should help here.

[Launchpad Janitor (janitor)]

[Expired for neutron because there has been no activity for 60 days.]