Flanneld not working in Kubernetes Multinode Example

Question

I am trying to follow this tutorial with two Vagrant instances: http://kubernetes.io/v1.0/docs/getting-started-guides/docker-multinode.html
After setting up master and worker node I tried to connect to the service's IP of a simple nginx-service from the master. But it looks like the kube-proxy cannot find the docker-container of the worker-node.

The virtual IP of the service and the container-ip respond well on the worker-node

That made me think of a misfunction of flanneld.
Does anybody know how I could track down this error?

Any help is appreciated!
Thanks in advance
Best, Johannes

Output of Kube-Proxy container

I1016 20:53:42.829290 1 proxysocket.go:130] Accepted TCP connection from 10.0.2.15:51774 to 10.0.2.15:40197
E1016 20:53:43.829575 1 proxysocket.go:99] Dial failed: dial tcp 10.1.12.3:80: i/o timeout
E1016 20:53:45.825473 1 proxysocket.go:99] Dial failed: dial tcp 10.1.12.3:80: no route to host
E1016 20:53:48.825556 1 proxysocket.go:99] Dial failed: dial tcp 10.1.12.3:80: no route to host
E1016 20:53:51.825627 1 proxysocket.go:99] Dial failed: dial tcp 10.1.12.3:80: no route to host
E1016 20:53:51.825710 1 proxysocket.go:133] Failed to connect to balancer: failed to connect to an endpoint.

the UDP port 8472 was blocked on our machines, vxlan of flanneld is using this port by default. After unblocking it their was no problem no more. nc -u helped me — johannes.schmidt.delaunay, Nov 02 '15 at 13:25

score 0 · Answer 1 · answered Oct 21 '15 at 12:21

0

This looks like https://github.com/kubernetes/kubernetes/issues/14426. Try upgrading flannel to 0.5.3.

answered Oct 21 '15 at 12:21

Filip

573
5
19

Flanneld not working in Kubernetes Multinode Example

1 Answers1