@aanm has been hitting this issue with the following developer commandline:

sudo service nfs start && ETCD_CLEAN=1 RELOAD=1 NWORKERS=1 NFS=1 K8S=1 contrib/vagrant/start.sh

==> cilium-k8s-node-2: Jan 23 15:34:54 cilium-k8s-node-2 cilium-agent[2744]: time="2018-01-23T15:34:54-08:00" level=debug msg="Cannot establish connection to local cilium instance" error="Get http://%2Fvar%2Frun%2Fcilium%2Fcilium.sock/v1beta/healthz: dial unix /var/run/cilium/cilium.sock: connect: no such file or directory

Sometimes it shows up when running like the above, sometimes not. The messages repeat for maybe up to 90 seconds, then it recovers.

The issue appears to be cilium-health complaining that it is unable to reach the cilium API via /var/run/cilium/cilium.sock. It's not clear why the cilium socket is sometimes unavailable on startup like this.