cardoe | So trying to steer the versioning of OpenStack Helm charts a little bit in https://review.opendev.org/c/openstack/openstack-helm/+/934013 (you'll have to expand the convo) | 00:51 |
---|---|---|
cardoe | Trying to come up with a solution where the tarball doesn't change after its built. | 00:51 |
cardoe | My suggestion right now is when something lands in ^ironic/.*$ path then a new tag can be made like "ironic-0.2.21" (assuming the prior was 0.2.20). Gerrit can see that new tag and build the tarball from there. | 00:52 |
cardoe | Right now there's a "helm-toolkit" library from openstack-helm-infra that triggers rebuilding all the tarballs in openstack-helm and puts a new copy of helm-toolkit into each of them at the current version. | 00:53 |
cardoe | This can lead to an assortment of breakages. Like labels are set by a function in helm-toolkit so that means everything has new labels. And you match resources by labels so surprise breakage. | 00:54 |
cardoe | The PTL is saying that tags are difficult with gerrit and will involve governance. Just wondering how painful that really will be? | 00:55 |
cardoe | At the end of the day I would really like things to be immutable when built. | 00:56 |
cardoe | I'm posting this here hoping that folks have some suggestions or ideas to get us past the impasse. | 00:56 |
clarkb | cardoe: minimally gerrit requires you sign the tag (the way our acls are typically set up) and then you push them thats all | 16:35 |
clarkb | cardoe: I suspect they're trying to say that all tagging also has to go through the release team's system which I don't think is true. THere are a couple of exceptions to that and "this isn't python software deliverables" make sense to me as another reasonable exception | 16:36 |
clarkb | I guess my point re tags and gerrit is they are no more difficult than any other git system using signed tags | 16:36 |
opendevreview | Merged openstack/election master: Fix the direct reminder template https://review.opendev.org/c/openstack/election/+/931714 | 18:18 |
clarkb | I've just discovered that you cannot pip install python-openstackclient without a full build toolchain on python3.10 or newer due to netifaces being a dependency | 19:12 |
clarkb | netifaces is old an unmaintained and doesn't have wheels for newer pythons. I'm also not sure why openstaclient would need it as a dependency in the first place | 19:13 |
cardoe | it's very minor | 19:18 |
cardoe | I also brought up the above in #openstack-infra and we had a long-ish chat there | 19:18 |
clarkb | it also looks like we've lost the ability to specify the clouds.yaml path or maybe it was never documented before and I've just lost the tribal knowledge | 19:19 |
cardoe | So I think the netifaces had a suggested removal. But it required something to land in global-requirements.txt as a replacement and it didn't | 19:21 |
cardoe | clarkb: def localhost_supports_ipv6(): """Determine whether the local host supports IPv6 | 19:26 |
cardoe | That's all that netifaces is used for now days. | 19:26 |
clarkb | but why does the client care? | 19:27 |
cardoe | It's only used in "get_default_access_ip()" | 19:28 |
cardoe | It uses netifaces to check if there's an interface with a functioning IPv6 default route and then returns the IPv6 address otherwise IPv4 | 19:29 |
clarkb | ya mordred pointed out in another channel it does that so that you can connect to VMs using the appropriate IP version since those instances may not have DNS so you can't rely on your resolver to feed you the appropriate ip | 19:36 |
clarkb | still seems like it should be doable to determine if ipv6 is valid without netifaces. Really the issue is you can't install python-openstackclient in a straightforward manner on modern python | 19:36 |
sean-k-mooney | for what its worth netifaces is kind of unmainteiend | 20:27 |
sean-k-mooney | and we have started removing it form some project and replacing some of its usage with funcitons in oslo.utils i think | 20:28 |
sean-k-mooney | so it proably should be droped form osc at somepoint | 20:28 |
clarkb | ya the functioanlity doesn't seem critical but also could be approximated some other way | 20:51 |
clarkb | (like maybe just a force ipv6/ force ipv4 flag on the cli) | 20:51 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!