Hi all,
Last years of my work on OpenStackSDK and CLI on one hand and being representative of a public cloud on the other hand made clear to me that not a single OpenStack cloud behaves like another one, doesn’t matter what you do and how hard you try. And sadly Interop is not helping here at all. The API calls you use to talk with one public cloud are guaranteed to work slightly differently on another one. This is clearly visible through the Storyboard of SDK, IRC and other offline discussions. Lot of customers are capable to work without issues with SDK/CLI on their cloud, while others report the same things are not working for them.
This is a relatively complex discussion and Clark reserved a slot in the TC meeting during PTG https://etherpad.opendev.org/p/tc-2023-1-ptg#L38 https://etherpad.opendev.org/p/tc-2023-1-ptg#L38 to describe his stake in the problem.
For a quite some time I am trying to follow an approach of having possibility to run functional tests of both SDK and CLI on real clouds and not only on devstack (on devstack things are working fine but it means as much as nothing at all in real life). On multiple occurrences I was asking for credentials and I got at the moment credentials for 2 public clouds. This already helped both to find issues in SDK (i.e. RADOSGW != Swift) as well as to point clouds on real issues on their side. It should be actually in the interest of those clouds to have some useful checks that give guarantee to their users SDK and CLI (ideally also Ansible/TF) are working fine. This is even a better form of interoperability check than anything else what we have now.
Since people in this ML are involved in maintaining resource donations to OpenDev I would like to ask you to consider allocating cloud credentials for the SDK/CLI team to ensure things are running smoothly for your users (pls do not mix with allocating real resources - we need only cloud credentials to test APIs and there are no static resources required). At the moment I maintain those credentials manually until we finally get post-review pipeline created in OpenStack tenant to fully automate those tests.
If you have questions or suggestions please feel free to post here, join SDK/CLI team during PTG on Friday 27.10 13:00-15:00 or contact me directly for clarifications.
Thanks a lot for support, Artem