On Mon, 2015-12-14 at 23:30 +0530, Kushal Das wrote:
Hi,
This email is to resume the discussion on putting Networkd as network stack in Fedora Cloud Atomic and base images. During Fedora 23 release we had some discussion on the same. I made couple of images [1], and also AMI on AWS [2], so that people can play around.
Major Hayden did some excellent write up [3] [4] [5] about Networkd in Fedora [6].
Last time when we discussed this issue, one of the major concern was QA cycle. We all know how much overloaded the QA team is. Now we have some level of automated testing (increasing regularly), and also a new volunteer group from the Cloud SIG to help by doing manual QA on the images. If we can make a list of things to be tested (or what all tests are currently being done) on the network stack for current Fedora releases, we all can help in, and make sure that the images with this new feature get proper testing.
The thing is that it's almost impossible to say "if we just run all X tests, we can guarantee everything is fine!" in real life, especially at the level of something as complex as an entire OS networking stack. It's simply an unavoidable fact of life that the more network configuration stacks we have in mass usage, the more likely it is that there will be problems. We already have the legacy network.service and NetworkManager, adding a third choice is kind of egregious.
It's an oldie, but a goodie:
https://www.redhat.com/archives/fedora-devel-list/2008-January/msg00861.html