Kubernetes etcd bootstrapping - couldn't find local name <name> in the initial cluster configuration

Question

I followed kelsey Hightower's setting up kubernetes the hard way tutorial kubernetes the hard way on AWS

Stuck on bootstrapping etcd on the first control plane node. There are 2 lines in the error message which concerns me:

1. The node is already initialized as member, starting as member..

• i have tried deleting the data directory while reinitializing but still same message

2. couldn't find local name in the intial cluster configuration

Is it picking up old configuration data from somewhere other than the data directory?

Answer 1

Well, it's labeled as hard :-D

Without knowing more troubleshooting steps that you've already taken, it's hard to know what to suggest for that first item, but the second one is almost certainly that its own name is not present in the --initial-cluster list. You'll also want to give etcd the full hostname as AWS's DNS returns it (ip-10-20-30-40.us-east-2.compute.internal), to save yourself future troubles when those etcd members attempt to look up their own name. That's almost certainly true of the Node names if you choose to use cloud-provider=aws when you get to that part, too

If this is the beginning of your kubernetes journey, then this is a great way to establish your future relationship with etcd, as it will be the source of unending pain and opaque messages. I would actually also encourage you to pause and go through their docs because having a very firm grasp on what etcd wants, how to debug it, how to recover from it, and how to drive etcdctl will pay dividends over your future kubernetes lifespan

Answer 2

There was a mismatch in the --initial-cluster list and the local name i was providing. Be sure to check the configuration carefully when following the tutorial.