Bugfix

• #14 fix docker/Client.getNetwork() bug causing all docker network events to mark the network as down, leading to clusterf route removal on container start

Version 0.5.2: bugfix

• Simplified config merging
• #13 Fix etcd 3.x compatibility
• Slightly improved reliability in case of slow etcd refreshes.
  In case of "late refresh" log messages, increase --etcd-ttl.

Major rewrite of config, clusterf and docker packages

Use CoW state structs instead of events.

• --config-source=etcd://localhost:2379/clusterf replaces -etcd-machines= -etcd-prefix=
• --config-source=file://... replaces -config-path=...
• clusterf-docker --route-* flags replace clusterf-ipvs -advertise-route-*, using the runtime network information from Docker
• clusterf-docker uses etcd TTLs with the new refresh operation for keepalive
• clusterf-docker sets weight=0 for stopping backends
• clusterf-ipvs fixes many bugs/issues
• clusterf-ipvs --flush is no longer required, syncs against existing state instead
• clusterf-ipvs --ipvs-elide omits empty services
• clusterf-docker --filter-routes= replaces -filter-etcd-routes
• clusterf-config displays the configuration state

Breaking etcd schema changes:

• /routes/* fields have changed. Prefix and Gateway are IPv6/IPv6 agnostic, and IPVSMethod is required unless filtering.

Test coverage is also much improved