Issue with /docs/getting-started-guides/centos/centos_manual_config/

[bostone]

This is a...

• Feature Request
• Bug Report

Problem:
After replicating all steps using 2 CentOS 7 VM there are 2 errors:

1. Executing "etcdctl mkdir /kube-centos/network" issues the following error: "etcd cluster has no published client endpoints. Try '--no-sync' if you want to access non-published client endpoints(http://127.0.0.1:2379,http://127.0.0.1:4001). Error: client: no endpoints available"
2. Unable to start flanneld service. It fails with this error: Unable to start service flanneld: Job for flanneld.service failed because a timeout was exceeded. See "systemctl status flanneld.service" and "journalctl -xe" for details. Checking specified files shows the same error: "Failed to retrieve network config: client: etcd cluster is unavailable or misconfigured"

Proposed Solution:

1. Adding "--no-sync" to command removed the message. The key is created
2. I am not able to come up with any solutions

Page to Update:
http://kubernetes.io/docs/getting-started-guides/centos/centos_manual_config/

Kubernetes v1.3.0

I run all the steps outlined in the manual except I don't add IPs into /etc/hosts but use fully qualified (pingable) hostnames instead. I did try to use IPs instead of hostnames to the same effect

[bostone]

Ok, I think I have a solution to my #2:
There is a mismatch between flanneld.service and /etc/sysconfig/flanneld configuration (which wasn't there before). You need to change keys in the flanneld configuration file to match these in flanneld.service. Namely - change FLANNEL_ETCD to FLANNEL_ETCD_ENDPOINTS and FLANNEL_ETCD_KEY to FLANNEL_ETCD_PREFIX in /etc/sysconfig/flanneld. After that flanneld.service starts with no problem on both master and minion(s) nodes

[a-mccarthy]

It looks like this was completed by #2133, closing