While looking at builder49 issues, I noticed that the centos7 container we are using to build the code is never refreshed. That kinda mean that we are pushing a container with a potential out of data base systems, which is maybe something we want to avoid as a general rule of thumb. I see that we upgrade the package: https://github.com/gluster/glusterd2/blob/master/extras/nightly-container/provision.yml#L7 But upgrading the package is not sufficient, since we could have added and/or removed package and/or added/removed configuration file in the container, change that wouldn't be covered by the "yum upgrade". Putting that on project-infrastructure, since that's a side effect on how we do build rather than the build script. "buildah from centos:7" seems the correct way to do it, but since we do not clean the already downloaded container, we keep using a old one.
This should be fixed now that we do a prune every time.