Easily manage etcd cluster with env vars

[gianrubio]

Exporting etcdctl environment variables, allow cluster managers to run etcdctl commands without providing arguments.

Currently when you logging in the etcd cluster you can't run etcdctl commands except you provide all the args.

$ etcdctl --cert-file=/etc/etcd2/ssl/etcd-client.pem --ca-file=/etc/etcd2/ssl/ca.pem --key-file=/etc/etcd2/ssl/etcd-client-key.pem  --endpoints=https://ip-10-0-0-5.eu-west-1.compute.internal:2379 cluster-health
member f5654f631de6e415 is healthy: got healthy result from https://ip-10-0-0-5.eu-west-1.compute.internal:2379
cluster is healthy

After this PR will be easy for run etcdctl commands

$ etcdctl cluster-health
member f5654f631de6e415 is healthy: got healthy result from https://ip-10-0-0-5.eu-west-1.compute.internal:2379
cluster is healthy

Also this is important on #49 and signaling cloudformation when the etcd cluster is healthy.

[codecov-io]

Current coverage is 54.72% (diff: 100%)

Merging #58 into master will not change coverage

@@             master        #58   diff @@
==========================================
  Files             4          4          
  Lines          1038       1038          
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
  Hits            568        568          
  Misses          388        388          
  Partials         82         82          

Powered by Codecov. Last update 7237203...a47d7d6

[mumoshu]

I'm just curious but may I ask why you've explicitly emptied ETCD_ENDPOINT here?
Just to note that we have to explicitly set it when running etcdctl?

[mumoshu]

I realized that this line is the "marker" for https://github.com/coreos/kube-aws/pull/58/files#diff-c88cb625705bf9711a2f576a1975cb72R34.
Doing this makes it possible to:

• Mention that we have to provided an actual value when running etcdctl
• Ensure two things:
  • Appropriate ETCDCTL_ENDPOINT value is always configured on starting/restarting etcd2.service, automatically.
  • Restarting etcd2.service doesn't break /etc/environment as we can see in https://github.com/coreos/kube-aws/pull/58/files#diff-c88cb625705bf9711a2f576a1975cb72R34
    • For example, echo ETCDCTL_ENDPOINT="https://$(hostname):2379/" >> /etc/environment would break, or needlessly bloat the environment file at least, when etcd2.service got restarted.

[mumoshu]

Note to self:

• This would make user-data much bigger than before... Oh, we now have the --s3-uri option to workaround that. No problem 😃
• This does add additional credentials we didn't have in etcd nodes before. Any security concerns for saving more credentials in etcd nodes? No.

[mumoshu]

@gianrubio I'm impressed with your work 👍

[mumoshu]

Thanks for the contribution! Merging.