Two days back we had a very productive meeting in the Fedora Atomic Working Group. This post is a summary of the meeting. You can find all the open issues of the working group in this Pagure repo. There were 14 people present at the meeting, which happens on every Wednesday 5PM UTC at the #fedora-meeting-1 channel on Freenode IRC server.
Fedora 26 change proposal ideas discussion
This topic was the first point of discussion in the meeting. Walters informed that he will continue working on the Openshift related items, mostly the installer, system containers etc, and also the rpm-ostree. I have started a thread on the mailing list about the change idea, we also decided to create a wiki page to capture all the ideas.
During the Fedora 25 release cycle, we marked couple of Autocloud tests as non-gating as they were present in the system for some time (we added the actual tests after we found the real issue). Now with Fedora 25 out, we decided to reopen the ticket and mark those tests back as gating tests. Means in future if they fail, Fedora 25 Atomic release will be blocked.
The suggestion of creating the rkt base image as release artifact in Fedora 26 cycle, brought out some interesting discussion about the working group. Dusty Mabe mentioned to fix the current issues first, and then only jump into the new world. If this means we will support rkt in the Atomic Host or not, was the other concern. My reaction was maybe not, as to decide that we need to first coordinate between many other teams. rkt is packaged in Fedora, you can just install it in a normal system by the following command.
$ sudo dnf install rkt -y
But this does not mean we will be able to add support for rkt building into OSBS, and Adam Miller reminded us that will take a major development effort. It is also not in the road map of the release-infrastructure team. My proposal is to have only the base image build officially for rkt, and then let our users consume that image. I will be digging more on this as suggested by Dusty, and report back to the working group.
Next, the discussion moved towards a number of technical debts the working group is carrying. One of the major issue (just before F25 release) was about missing Atomic builds, but we managed to fix the issue on time. Jason Brooks commented that this release went much more promptly, and we are making a progress in that :) Few other points from the discussion were
- Whether the Server working group agreed to maintain the Cloud Base image?
- Having ancient k8s is a problem.
- We will start having official Fedora containers very soon.
Then the discussion moved to documentation, the biggest pain point of the working group in my mind. For any project, documentation can define if it will become a success or not. Users will move on unless we can provide clearly written instructions (which actually works). For the Atomic Working Group, the major problem is not enough writers. After the discussion in the last Cloud FAD in June, we managed to dig through old wiki pages. Trishna Guha is helping us to move them under this repo. The docs are staying live on https://fedoracloud.rtfd.io. I have sent out another reminder about this effort to the mailing list. If you think of any example which can help others, please write it down, and send in a pull request. It is perfectly okay if you publish it in any other document format. We will help you out to convert it into the correct format.
You can read the full meeting log here.