I'm running 10 nodes and about 160 pods in our cluster. The cluster is up for a month and suddenly this error happened frequently when making request among services in cluster

"response": {
    "err": {
      "code": "ENOTFOUND",
      "errno": "ENOTFOUND",
      "syscall": "getaddrinfo",
      "hostname": "svc-api-mapper",
      "host": "svc-api-mapper",
      "port": "80"
    }
  },
  "attempt": 0

the pods selected by the service are up and running, there was no event indicating that pods are up and down but somehow the service name svc-api-mapper cannot be resolved.

$ kubectl describe se svc-api-mapper
Name:           svc-api-mapper
Namespace:      production
Labels:         app=workflow-koa-api-mapper
Selector:       app=workflow-koa-api-mapper
Type:           LoadBalancer
IP:             10.3.0.33
Port:           <unnamed>   80/TCP
NodePort:       <unnamed>   30106/TCP
Endpoints:      10.2.8.126:80,10.2.80.248:80
Session Affinity:   None
No events.

When I looked at the skydns logs, there were a lot of errors that printed every second: 2015/11/22 05:40:00 skydns: can not forward, name too short (less than 2 labels): svc-rethinkdb-driver.

Notice that the ENOTFOUND error doesn't always happen, just frequently. Is it related to performance issue?

I'm using kubernetes 1.0.1 in CoreOS cluster 773.1.0