Don't respond to DHCP requests when unable to access etcd

[emmaly]

If etcd is down, do NOT reply to any DHCP requests. And certainly don't issue a DHCPNAK due to the etcd cluster downtime. Just pretend to not exist and wait for etcd to come back online, instead.
Should DNS act the same way? I think it should try to run in read-only mode with etcd if available and serve out of that (do we care about expiring/expired TTLs at that point? If yes, how/etc)

---

Note: This issue was migrated from https://bitbucket.org/dustywilson/netcore/issues/2 originally created on 2015-01-31 by Dusty Wilson. It was not evaluated in any way for freshness and validity.

[emmaly]

I think if etcd is unavailable and the netcore instance is already running (because it can't start running if etcd is unavailable), DHCP should just stop responding altogether and DNS should blindly forward out to another known netcore instance (if possible) or forward all queries out to upstream (if possible).