[OVN] Regression when using ML2/OVN with Neutron DHCP agent

[Lucas Alvares Gomes]

In an attempt to minimize the number of connections to the OVSDBs the ML2/OVN driver no longer connects to the OVN dbs if the Neutron worker is a RpcWorker [0].

This introduced a regression where when ML2/OVN is deployed along with the Neutron DHCP server (e.g for baremetal provisioning) the DHCP agent is no longer able to create the DHCP port and fails with the following error message:

2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers Traceback (most recent call last):
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers File "/usr/lib/python3.6/site-packages/neutron/plugins/ml2/managers.py", line 477, in _call_on_drivers
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers getattr(driver.obj, method_name)(context)
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers File "/usr/lib/python3.6/site-packages/networking_ovn/ml2/mech_driver.py", line 551, in create_port_precommit
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers if self._is_port_provisioning_required(port, context.host):
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers File "/usr/lib/python3.6/site-packages/networking_ovn/ml2/mech_driver.py", line 589, in _is_port_provisioning_required
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers if not self._sb_ovn.chassis_exists(host):
2021-07-13 13:36:18.116 17 ERROR neutron.plugins.ml2.managers AttributeError: 'NoneType' object has no attribute 'chassis_exists'

See full traceback at: http://paste.openstack.org/show/807435/

[0] https://opendev.org/openstack/neutron/commit/c4007b0833111a25d24f597161d39ee9ccd37189

[Rodolfo Alonso (rodolfo-alonso-hernandez)]

Hello:

In a "normal" deployment (without baremetal chassis) we don't use DHCP agents; this is why, as you commented, we removed the OVSDBs connectivity from other classes than WorkerService and MaintenanceWorker.

In this specific case, I think https://review.opendev.org/c/openstack/neutron/+/795781 is what you are looking for. This patch will allow you to define more server classes (RpcWorker in your case) that could have OVN IDL connectivity.

Does it work for you?

Regards.

[Rodolfo Alonso (rodolfo-alonso-hernandez)]

without baremetal servers* (not chassis)

[Rodolfo Alonso (rodolfo-alonso-hernandez)]

Related bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1972774

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/neutron/+/800679

[Lucas Alvares Gomes (lucasagomes)]

Hi Rodolfo,

Thanks for pointing that out. Yes that option does work but, I don't think it should be the final fix for this.

For one, it's not backportable (since it adds a config option). And also, we are now by default breaking a use-case that worked before, this would require operators to add support and change their scripts/tools to accommodate to that option to fix something that already worked.

[Bartosz Bezak (bbezak)]

there were similar bugs before:
https://bugs.launchpad.net/neutron/+bug/1929633
https://bugs.launchpad.net/neutron/+bug/1929197
https://bugs.launchpad.net/neutron/+bug/1928794

addressed with https://review.opendev.org/c/openstack/neutron/+/781555/

[Lucas Alvares Gomes (lucasagomes)]

Hi Bartosz,

Thanks much for the pointers. I can see how this bug is duplicated with [0] indeed. I will mark it as such.

For the patch [1] I do not think it solves the problem because RpcWorkers (by default due to the new configuration option) will not connect to the OVN databases (this is true for stable branches where the config option is not present).

[0] https://bugs.launchpad.net/neutron/+bug/1929633
[1] https://review.opendev.org/c/openstack/neutron/+/781555/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/wallaby
Review: https://review.opendev.org/c/openstack/neutron/+/800780

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/victoria
Review: https://review.opendev.org/c/openstack/neutron/+/800781

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/ussuri
Review: https://review.opendev.org/c/openstack/neutron/+/800782

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Alvares Gomes <email address hidden>" on branch: stable/ussuri
Review: https://review.opendev.org/c/openstack/neutron/+/800782
Reason: Will cherry-pick from wallaby

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Alvares Gomes <email address hidden>" on branch: stable/victoria
Review: https://review.opendev.org/c/openstack/neutron/+/800781
Reason: Will cherry-pick from wallaby

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/ussuri
Review: https://review.opendev.org/c/openstack/neutron/+/800783

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/victoria
Review: https://review.opendev.org/c/openstack/neutron/+/800784

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Alvares Gomes <email address hidden>" on branch: stable/wallaby
Review: https://review.opendev.org/c/openstack/neutron/+/800780
Reason: We've changed the master branch approach and now the patch is backportable

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Alvares Gomes <email address hidden>" on branch: stable/victoria
Review: https://review.opendev.org/c/openstack/neutron/+/800784
Reason: We've changed the master branch approach and now the patch is backportable

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Alvares Gomes <email address hidden>" on branch: stable/ussuri
Review: https://review.opendev.org/c/openstack/neutron/+/800783
Reason: We've changed the master branch approach and now the patch is backportable