Test Template

• test ubercontainer failover onto another host when the original host fails (see below)
  • pods should keep running on other hosts
  • pods on killed host should be restarted (if backed by rc)
  • dns resolution should keep working: in busybox container nslookup kube-ui.kube-system.svc.cluster.local
  • dns resolution should work for new services: launch nginx pod and service. Test in busybox container nslookup nginx.default.svc.cluster.local.
• deploy examples/guestbook. Try posting something in the web UI.
• execute dcos kubectl version
• do 2h+ soak testing w/ hack/resizeFrontend.sh
• smoke test Mesos-DNS via a busybox test container executing nslookup kubernetes.mesos.
• smoke test roles support using a simple nginx pod (see [1]). Install with dcos kubectl create -f nginxpub.yaml, wait ~1min. and open the link "Master IP addresses:" available in CCM. One should see the nginx landing page.

ubercontainer failover

Find out which slave ID runs the kubernetes ubercontainer in the DCOS UI, then ssh into the node:

$ dcos node ssh --master-proxy --slave=xxxxxx-yyyy-4cab-b1fb-3c92a11cdd2d-S1
$ sudo systemctl stop dcos-mesos-slave.service

[1] Sample roles enabled public nginx pod nginxpub.yaml:

apiVersion: v1
kind: Pod
metadata:
  name: nginxpub
  annotations:
    k8s.mesosphere.io/roles: "slave_public"
  labels:
    app: nginxpub
spec:
  containers:
  - name: nginxpub
    image: nginx
    ports:
    - containerPort: 80
      hostPort: 80