[09:28:06] I hold my hand up that the question about was probably brought about because I wrote this task https://phabricator.wikimedia.org/T303381 and because it took me six months to get the upgrade to DataHub done: https://phabricator.wikimedia.org/T329514 [09:28:25] ...about alpine... [09:31:22] However, I didn't really push for alpine adoption and didn't really state that there was an operational need. It was just one possible solution of how I could have got around the issues with the build process. [09:32:03] I've now effectively un-forked our datahub build scripts from upstream and made significant progress with how it builds with blubber on Debian from pristine sources, so I'm much happier with things as they are. [10:29:05] <_joe_> btullis: ack, thanks for the context [10:29:48] <_joe_> btw, I was thinking that we need to improve that pipeline [10:29:58] <_joe_> it would be great if we managed to run it from gitlab's CI [10:30:23] <_joe_> we will need to reach out to releng to understand how I need to modify docker-pkg to work in a CI trusted worker [10:30:31] <_joe_> and the SRE collab team too [10:30:59] <_joe_> but I'd love to have a CI => build with docker-pkg, including running its tests, without publishing [10:31:10] <_joe_> and on merge, also publish to our docker registry [10:33:53] _joe_: Yes, I've already got a ticket to migrate the datahub pipeline to GitLab CI. I was waiting for some input on it from releng, but to be honest I think I may just go ahead and do it. This one is all blubber, so no need for docker-pkg. [10:34:03] https://phabricator.wikimedia.org/T341194 [10:34:28] <_joe_> btullis: oh yes, I guess it's just the base images that you'll need [10:35:14] But I agree, getting the base-images and production-images onto gitlab would be a win. [10:36:08] btullis: oh, that's useful info, thanks. Happy to hear that you solved the problem already. I think we still wanna have the larger discussion regarding alpine images, and document pros/cons and output of that discussion so we can refer to it in the future. [10:36:22] +1 on the gitlab CI stuff btw [10:41:22] akosiaris: Thanks. Yes, I agree. I've also started drafting a page here about evaluating the DataHub helm charts and whether we should carry on with our custom chart or switch to upstream. https://wikitech.wikimedia.org/wiki/Helm/Upstream_Charts/Datahub [10:42:27] I deliberately haven't linked it from the parent page yet, because I'm still undecided myself one whether we should even undertake the review. [12:01:49] thanks for that page. I guess it might make sense to run the process once with 2 people with a "known" subject. Datahub's situation might prove to be an advantage in this cause. We are still undecided as I understand it and that means our bias is somewhat limited (hopefully). So, we might kill 2 birds with 1 stone and gain some clarity on Datahub [12:01:50] and some experience with the process.