Lab 3.4: Can't reach to ClusterIP and Endpoint with curl from master to worker node

cumulus

In Exercise 3.4 Step 19-20, the document instruct me to run a nginx on worker node and then
curl $clusterIP:80
curl $endpoint
on MASTER node, with tcpdump tunl0 on worker.
In my lab on GCP, I can get nginx default page with curl on WORKER node for both addresses. However if I run curl on master node, it doesn't return (waiting timeout). The odd thing is that tcpdump on worker node print nothing.
kubectl works fine to get/create/delete/describe objects on the cluster, for both of nodes.
Any suggestions to troubleshoot?

chrispokorni

Hi,

Feel free to use the search feature of the forum, as most issues have been reported and solutioned already. Below are only a few discussions on this very same topic, but there are others:

https://forum.linuxfoundation.org/discussion/856313/gcp-cannot-ping-instances#latest

https://forum.linuxfoundation.org/discussion/855983/lab-3-now-on-gcp-cant-get-nodes-to-curl-nginx#latest

I also suggest reading carefully the overview section of lab 3.1 because it presents helpful configuration options for inter-node networking.

Regards,
-Chris

cumulus

I'm surprised at my blindness on the text marked as "Very important".
Thanks Chris.