Hey Lajos, Slawek. firstly I am truly very sorry about not responding to your kind message in any way until just now. I myself simply expected to be able to reproduce the issue quicker and postponed any response way too long. In any case: We were now able to reproduce the issue on DevStack with xena or master, see me comment at the end of this message. On 13/01/2022 09:34, Lajos Katona wrote:
Thanks for highlighting your issue with Linuxbridge driver and VPNaaS. I would like to ask you to consider the following, which will not give you an answer or solution to your problem but perhaps helps to understand and accept if any specific issue has not received the needed attention from the community. Basically it is the core team (and some enthusiasts around it) who do bug triaging in Neutron, they are experienced and willing to provide answers, solutions. The possible combinations of drivers, extensions, backends, deployment tools is huge, and there are combinations which are not well covered in the current community, like linuxbridge is not that used (OVS and perhaps OVN are the top tested and deployed, at least from latest user surveys). So it can happen that in the current core team nobody has the special knowledge and tools (test environment, hardware...) to debug a specific issue.
I understand and thank you for your patience in explaining this (likely not for the first time). I am already tracking the progress happening in regards to OVN, but unfortunately support for VPNaaS is still not available there, but there is a pending change at https://review.opendev.org/c/openstack/neutron-vpnaas/+/765353. On 13/01/2022 09:18, Slawek Kaplonski wrote:
I don't just want to shout louder and certainly everybody wants their issue to be looked at and fixed first. But I just noticed that my bug report received no reply or confirmation in the last three months (apart from the tag "vpnaas" being added), while other issues were triaged quickly and were consequently added to this deputy report ... The problem with neutron-vpnaas is that there is almost no one who is maintining that project currently and probably because of that nobody works on the bug which You reported. If You are using that project and are interested in maintaining it, patches are always welcome:)
Not to push back on anything your said, it's just a bit frustrating sometimes that there seems to be no explicit deprecation and then EoL path for certain backends / drivers. To me this would be totally understandable as a way to free development time and bug hunting efforts. And at the same time this means providing some strong guidance to the userbase what to avoid or actively moving away from. The most prominent example to me, as our deployment was affected by this as well, is using PostgreSQL as DBMS. While there is this message, https://governance.openstack.org/tc/resolutions/20170613-postgresql-status.h..., explaining that there will be less focus on PostgreSQL, it would actually be much better to simply deprecate support for this DBMS and then clearly state this for the next releases. With a state of "supported, but not widely used", users will just have a hard time getting their reported bugs validated and fixed. Even reviewers for sent in patches are hard to find, see e.g. https://review.opendev.org/c/openstack/designate/+/668493. But even if an issue was found and fixed, as https://review.opendev.org/c/openstack/glance/+/820247, a similar issue is just around the corner, as this configuration is just not actively tested and validated. Looking at Neutron there are lots of tables and comparisons. But I am not convinced the feature tables and complex interdependencies at * https://docs.openstack.org/neutron/latest/feature_classification/general_fea... * https://docs.openstack.org/neutron/latest/admin/config-ml2.html#id5 * https://docs.openstack.org/neutron/latest/ovn/gaps.html help a new Neutron user to make the right choices. I already somewhat read between the lines that OVS support could be dropped in favor of OVN at some point - like RedHat recommends and explains here: https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.... What I am trying to say is that it would be ok to drop support for something at some point to free time and effort to focus on more common configurations and making them work as best they can. But this also includes making active recommendations and getting the missing features tackled (as in VPNaaS for OVN) so people can actually be asked to migrate away from a EoL config. On 13/01/2022 09:34, Lajos Katona wrote:
Neutron maintains a list of "lieutenants", to make easier to contact the right person: https://docs.openstack.org/neutron/latest/contributor/policies/neutron-teams... For VPNaaS the people are not active anymore in the community. In such difficult to debug situation it is really helpful for the community if you can test the issue on current master code (it is possible that the issue happens only on older branches) and with tools that are available for most of us, like simple devstack.
My colleague Niklas Schwarz did setup two devstack nodes and documented their config as well as required steps setting up VPNs to then trigger the reported duplicate iptables issue, see https://bugs.launchpad.net/neutron/+bug/1943449/comments/4. Please kindly take a look and let me know if there is any more details we need to provide. But since we have this issue reproducible in devstack I hope you'll be able to narrow it down. Regards and thanks again for your patience, Christian