In this issue we will try to track the work we are doing to modify our templates in openstack/magnum to deploy kubernetes (etcd, flannel) in system containers.
At the moment, we run 1. etcd, flannel, kubelet and kube-apiserver from the binaries includes in fedora-atomic 2. kube-proxy, kube-scheduler, kube-controller-manager as static pods 3. core-dns, kubernetes-dashboard as deployments
The goal is to run all components of 1. in system containers and be able to change version easily and get the newest faster. We will investigate if there are benefits to run 2. as system-containers too.
We (magnum and atomic-wg) could leverage magnum to test the newest packages of kube/etcd/flannel. Magnum's CI will run soon on CICO [1].
Magnum's configuration scripts: http://git.openstack.org/cgit/openstack/magnum/tree/magnum/drivers/common/templates/kubernetes/fragments
[1] https://bugs.centos.org/view.php?id=13192
https://review.openstack.org/#/c/487357/
In this patch we pull the apiserver and kubelet from our project repo in dockerhub [2].
I build the images myself from atomic-system-containers from this [1] commit and tagged them with v1.7.1
[1] https://github.com/projectatomic/atomic-system-containers/commit/de1219105dd38859e520327a8f4e4db47a0f5c3c [2] https://hub.docker.com/r/openstackmagnum/kubernetes-kubelet/tags/ https://hub.docker.com/r/openstackmagnum/kubernetes-apiserver/tags/
Tested locally, works fine.
Use atomic containers for kubelet & apiserver https://review.openstack.org/#/c/487357/ Launch k8s scheduler & controller-manager as system containers https://review.openstack.org/#/c/487943/ Launch kube-proxy as a system container https://review.openstack.org/#/c/488443/
We can close this issue, the above patches are merged.
OpenStack/Magnum Pike 5.x is deploying kubernetes in system containers. https://docs.openstack.org/releasenotes/magnum/pike.html
Login to comment on this ticket.