Max Gautier
Max Gautier
I think we need to slip a backport of #12044 as well ; external etcd setup seems broken in some cases without it.
> Blocking issue [#12061](https://github.com/kubernetes-sigs/kubespray/issues/12061) This won't be backported: while a better default is good, this does not fix a bug. @arusa You mean on release-2.27 ? Would you mind expanding...
Thanks ! I wonder if a partial backport could fix the issue 🤔 @chadswen do you think we could backport 94deb52121357221a550395f801896e956482726 only (part of #12015 ), or maybe the whole...
No it has been adressed in #12192 and #12221, we should be good.
> [@VannTen](https://github.com/VannTen) The release-note is ready, is there anything else you would like to add or suggest? Nope don't have anything , sorry for the delay.
Hum, the problem is that this takes twice the time of the current job, and since it's required for all jobs, it's not an insignificant impact on the pipeline runtimes...
Ideally I think we should switch to build a base image in CI and rely on caching pip packages and collection, but that's needs some work and I'm not sure...
> If it can be executed in less than 2 minutes, it should be cached (I guess?), but the original task may take more than 2 minutes to run. I'm...
Cache looks much more efficient that Kaniko, from what I see by restarting. I'll try another test PR based on this one to see if a different commit SHA kills...
/lgtm Thanks ! Should make the image build much more annoying for pipelines times.