Dr. Stefan Schimanski
Dr. Stefan Schimanski
A 100 lines sound like reasonable large. If it doesn't make the code completely ugly, I would try to factor that out.
After looking at the code, the only dependency to the node controller are the kubeclient and some event sink. Looks sensible to factor out,
kubectl exec through insecure ports does not work. It needs SPDY on top of https.
@jdef this is more like an epic and probably too much for the MVP. We have https://github.com/mesosphere/kubernetes-mesos/issues/573, https://github.com/mesosphere/kubernetes-mesos/issues/410 and https://github.com/mesosphere/kubernetes-mesos/issues/353 labeled with MVP, not the etcd-mesos topic.
Is it a typo with xx.com and hp.com: `error connecting to proxy http://web-proxy.corp.xx.com:8080/: dial tcp: lookup web-proxy.corp.hp.com` ?
Do the other components like the scheduler and the kubelets work behind the proxy? They should use the same mechanisms to communicate with the apiserver.
I see a DNS error and `403 Forbidden`. Where are those from?
`dial tcp: lookup web-proxy.corp.xx.com on 16.155.255.10:53: read udp 10.100.3.41:36156->16.155.255.10:53: i/o timeout`, port 53 is DNS usually.
I guess you have maximal logs already, i.e. log-level 9?
@jdef: do we have a Jira for this?