0

The way it is single node kubernetes on a OpenStack VM is:

VMs IP : 10.120.20.227 (from outside)
etcd version 3.0.16
kubectl --version 1.5.7
Flannel version 0.6.1

When I ssh in to the machine I could see the ip 192.168.0.5 So etcd service is running on 192.168.0.5 I could access every application launched on in the VMs from VM its self. But from out side openstack cluster I am unable to access the application using VMs Public IP.

The kube-proxy errors are

May 22 18:38:16 poc-desktop kube-proxy[1246]: I0522 18:38:16.293261    1246 server.go:215] Using iptables Proxier.  
May 22 18:38:16 poc-desktop kube-proxy[1246]: W0522 18:38:16.293629    1246 server.go:468] Failed to retrieve node info: Get http://192.168.0.5:8080/api/v1/nodes/poc-desktop: dial tcp 192.168.0.5:8
May 22 18:38:16 poc-desktop kube-proxy[1246]: W0522 18:38:16.293761    1246 proxier.go:249] invalid nodeIP, initialize kube-proxy with 127.0.0.1 as nodeIP
May 22 18:38:16 poc-desktop kube-proxy[1246]: W0522 18:38:16.293773    1246 proxier.go:254] clusterCIDR not specified, unable to distinguish between internal and external traffic

If Iaunch any webservice on VM on any random port I access the web app. But if I launch application using kubectl I am unable to access from other areas Does this require any special routing ? or something is wrong with kube-proxy ?

whites11
  • 12,008
  • 3
  • 36
  • 53
thunder hill
  • 1
  • 1

2 Answers2

1

I tried to similar thing. So I did setup a single node k8s cluster using kubeadm. To setup k8s cluster using kubeadm read more here.

And while starting the cluster I took care that I am exposing the public IP. Now if you look at this machines' default gateway address which is 172.17.133.24.

$ ip a sh eth0 | grep inet
    inet 172.17.133.24/24 brd 172.17.133.255 scope global dynamic eth0

But this IP address is internal to the machine and you cannot reach this IP address from outside. And on the OpenStack console I could see one more address associated with the instance which is what I can ping this machine on(from outside) and it is 10.3.8.95.

Now if you see while starting the cluster using kubeadm I used this IP address of 10 series.

# kubeadm init --skip-preflight-checks --apiserver-advertise-address=10.3.8.95                                                              
...
[addons] Created essential addon: kube-proxy
[addons] Created essential addon: kube-dns

Your Kubernetes master has initialized successfully!

To start using your cluster, you need to run (as a regular user):

  sudo cp /etc/kubernetes/admin.conf $HOME/
  sudo chown $(id -u):$(id -g) $HOME/admin.conf
  export KUBECONFIG=$HOME/admin.conf

...

Now when I started an app on k8s I exposed it via a service and then changed the service's type from ClusterIP to NodePort.

kubectl run web --image centos/httpd
kubectl expose deployment web --port 80
kubectl edit svc web

Changed the service's type to NodePort in the last command.

Now find the port on the machine this svc is exposed on:

$ kubectl get svc web
NAME      CLUSTER-IP      EXTERNAL-IP   PORT(S)        AGE
web       10.105.242.27   <nodes>       80:31628/TCP   19m

In above command you can see that the service web is exposed on two ports 80 and 31628. When you expose a service as NodePort it is exposed on a random port which is more than 30000.

Now to access this port from outside I have created a security group in OpenStack and allowerd tcp ports from 30000 to 60000. And added this security group to the machine.

Now from my laptop I can curl on the machine

$ curl 10.3.8.95:31628
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd"><html><head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
                <title>Apache HTTP Server Test Page powered by CentOS</title>
                <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
...
surajd
  • 1,627
  • 13
  • 13
0

Checkout your security group, may your pack from outside be dropped.

luke
  • 36
  • 4