0

I'm new to K8s and struggling with some problems with kube-proxy. I have a service that needs to expose the deployment. Everything seems correct but it does not expose it. I found the problem in kube-proxy logs.

Unknown proxy mode "", assuming iptables proxy

No output for the command iptables-save | grep hostnames

I thought adding --proxy-mode into the kube-proxy manifest would resolve my issue and couldn't find it.

I also tried to check the config file with this command:

ps -ef | grep kube-proxy

Output:

root      1423 24476  0 12:12 pts/0    00:00:00 grep --color=auto kube-proxy
root      8424  8392  0 11:45 ?        00:00:00 /usr/local/bin/kube-proxy --config=/var/lib/kube-proxy/config.conf --hostname-override=
ip-10-0-0-100.us-west-2.compute.internal
root@ip-10-0-0-100:~# find /var/lib/kube-proxy/config.conf
find: ‘/var/lib/kube-proxy/config.conf’: No such file or directory
root@ip-10-0-0-100:~# find  /usr/local/bin/kube-proxy
find: ‘/usr/local/bin/kube-proxy’: No such file or directory

Any leads to help me find my way would highly be appreciated. Thanks!

Wytrzymały Wiktor
  • 11,492
  • 5
  • 29
  • 37
cosmos-1905-14
  • 783
  • 2
  • 12
  • 23
  • Could you share details about how you run kube-proxy? – whites11 Apr 29 '21 at 05:48
  • I am doing a practice test for CKA exam. So I'm troubleshooting the issue. I can check something you want if you know what might cause this issue? @whites11 – cosmos-1905-14 Apr 29 '21 at 12:59
  • Well, to begin with "Unknown proxy mode" is not an error just a warning. My guess is that the path of kube-proxy in the manifest that runs it is wrong and you need to adjust it. You need to find how it is deployed – whites11 Apr 30 '21 at 06:15
  • @whites11 that was my question actually. I knew I could resolve the issue. I just cant find how to find the proper manifest for kube-proxy. As I said this is a practical lab exam question so it does not provide how kube-proxy is deployed. – cosmos-1905-14 Apr 30 '21 at 13:53
  • Normally kube-proxy is deployed as a daemonset in kube-system namespace. If you don't have to fix a deployment pipeline but just the status quo of the cluster try checking if there is such daemonset – whites11 Apr 30 '21 at 14:18
  • I added a comment about how kube-proxy is usually deployed. Did you check that? There is no progress I can make with the information I currently have – whites11 May 06 '21 at 10:08
  • Thank you for following up @WytrzymałyWiktor I have figured it out. The problem was a network policy. – cosmos-1905-14 May 20 '21 at 03:14

1 Answers1

0

One of the reason could be the network plugin not being installed out of many possibilities.

In my case, when I received the below error, I wanted to check for the installed CNI plugin.

Unknown proxy mode "", assuming iptables proxy

  • Check installed CNI plugin with below command,

    $ ls /etc/cni/net.d

  • Suppose, it's a weave-net. Then we must look for the objects created for this plug-in with k get all -n kube-system. In my case, I didn't find any objects created for the weave-net. Thus I just installed weave with the below command,

    $ kubectl apply -f "https://cloud.weave.works/k8s/net?k8s-version=$(kubectl version | base64 | tr -d '\n')"

One more thing to keep in mind,

/var/lib/kube-proxy/config.conf is the mounted location inside the kube-proxy pod. Check the details of ConfigMap named kube-proxy.

kubectl describe configmap kube-proxy -n kube-system
Jay Modi
  • 3,161
  • 4
  • 35
  • 52