kid
kid copied to clipboard
Kubernetes 1.6 versions don't seem to work correctly.
I've tried setting KUBERNETES_VERSION=1.6.2 but although the correct hyperkube version is pulled (gcr.io/google_containers/hyperkube-amd64:v1.6.2) it fails to start correctly with Docker reporting the container as stopped.
Googling for "kubelet 1.6 in docker" there's a little bit of traffic that seems to talk of cgroup issues https://github.com/kubernetes/kubernetes/issues/43704 but I've been unable to make any headway with any of the comments mentioned in the link, e.g.: "So I think we have found a work around now, adding --cgroups-per-qos=false --enforce-node-allocatable="" solves the problem." "I already have -v /sys:/sys:rw -v /sys/fs/cgroup:/sys/fs/cgroup:rw, or is that not enough or somehow not the right thing?" "So in release#306 it's been mentioned that --cgroup-driver=systemd is another working solution for RHEL"
I tried tweaking the docker run command in the kid "start_kubernetes" function as follows:
docker run \
--name=kubelet \
--volume=/:/rootfs:ro \
--volume=/sys:/sys:rw \
--volume /sys/fs/cgroup:/sys/fs/cgroup:rw \
--volume=/var/lib/docker/:/var/lib/docker:rw \
--volume=/var/lib/kubelet/:/var/lib/kubelet:rw,shared \
--volume=/var/run:/var/run:rw \
--net=host \
--pid=host \
--privileged=true \
-d \
gcr.io/google_containers/hyperkube-amd64:v${kubernetes_version} \
/hyperkube kubelet \
--containerized \
--cgroups-per-qos=false --enforce-node-allocatable="" \
--cgroup-driver=systemd \
--hostname-override="127.0.0.1" \
--address="0.0.0.0" \
--api-servers=http://localhost:${kubernetes_api_port} \
--config=/etc/kubernetes/manifests \
--cluster-dns=$DNS_SERVER_IP \
--cluster-domain=$DNS_DOMAIN \
--allow-privileged=true --v=2 \
> /dev/null
To include some of the stuff mentioned, but no joy.
Have you managed to get k8s 1.6.x working? Any thoughts on what I might be missing?
FWIW I'm running Docker version 17.05.0-ce, build 89658be on Linux Mint (Ubuntu based distro)
MTIA.