Dave Huseby <dhuseby@...>
After receiving feedback from many of you that you didn't think the CI/CD committee should be private, the CA team has decided to open it all up. I am keeping the one page with all of the outside CI/CD costs private to protect the confidentiality I promised when gathering expense information from companies currently paying to run Hyperledger CI/CD pipelines.
The CI/CD space can be found here:
You can see all of the ongoing research and ideas as well as the meeting minutes from our meetings. If you would like to participate, I am still looking for people to join the committee to help us work out solutions.
The biggest problem we face now is that many of the CI/CD pipelines for Hyperledger projects use docker-in-docker and docker-compose to first build their CI docker image and test it before running it to build the project code and test that. The current idea of airlifting existing pipelines onto a Hypereldger Kube cluster is blocked by the current docker-in-docker requirement.
Thanks to Mark, we have been thinking about how to get away from docker-in-docker since last October at the Montreal meetup. RedHat has a good solution called Buildah and there are a few others. We could really use help examining the different CI pipelines for the different projects to gauge the level of impact switching from docker-in-docker to something like Buildah would cause.
Kube is not a silver bullet. So the end recommendation from the committee should account for the cost moving all of the projects to a Kube cluster, in terms of estimated work hours and risk due to change.
If you can spare the cycles and know your project's CI pipeline fairly well, we could use your help.