"address already in use" integration test flake

[smarterclayton]

I0409 14:17:42.800392    2593 master.go:439] Started Kubernetes API at 127.0.0.1:60623/api/v1beta1
I0409 14:17:42.800412    2593 master.go:439] Started Kubernetes API at 127.0.0.1:60623/api/v1beta2
I0409 14:17:42.800416    2593 master.go:439] Started Kubernetes API at 127.0.0.1:60623/api/v1beta3
I0409 14:17:42.800419    2593 master.go:439] Started OpenShift API at 127.0.0.1:60623/osapi/v1beta1
I0409 14:17:42.800422    2593 master.go:439] Started OAuth2 API at 127.0.0.1:60623/oauth
I0409 14:17:42.800425    2593 master.go:439] Started login server at 127.0.0.1:60623/login
I0409 14:17:42.800428    2593 master.go:439] Started OpenShift UI 127.0.0.1:60623/console/
I0409 14:17:42.800432    2593 master.go:439] Started Swagger Schema API at 127.0.0.1:60623/swaggerapi/
I0409 14:17:42.823281    2593 net.go:39] Got error &net.OpError{Op:"dial", Net:"tcp", Addr:(*net.TCPAddr)(0xc208db9440), Err:0x3d}, trying again: "127.0.0.1:60623"
I0409 14:17:42.948517    2593 master.go:496] No master policy found.  Creating bootstrap policy based on: /var/folders/zw/j6656mbn1rv4fnqqdggss8h80000gn/T/openshift-integration/policy/policy.json
I0409 14:17:43.045505    2593 factory.go:79] creating scheduler from algorithm provider 'DefaultProvider'
I0409 14:17:43.045516    2593 factory.go:129] creating scheduler with fit predicates 'map[HostName:{} PodFitsPorts:{} PodFitsResources:{} NoDiskConflict:{} MatchNodeSelector:{}]' and priority functions 'map[LeastRequestedPriority:{} ServiceSpreadingPriority:{}]
I0409 14:17:43.046122    2593 master.go:117] Started Kubernetes Scheduler
I0409 14:17:43.046134    2593 master.go:94] Started Kubernetes Replication Manager
I0409 14:17:43.046138    2593 master.go:102] Started Kubernetes Endpoint Controller
I0409 14:17:43.262383    2593 master.go:152] Started Kubernetes Minion Controller
I0409 14:17:43.262416    2593 master.go:87] Started Kubernetes Namespace Manager
I0409 14:17:43.262429    2593 start_master.go:363] Using images from "openshift/origin-<component>:"
W0409 14:17:43.262530    2593 master.go:605] Binding DNS on port 60624 instead of 53 (you may need to run as root and update your config), using 127.0.0.1:60624 which will not resolve from all locations
2015/04/09 14:17:43 skydns: ready for queries on local. for tcp://127.0.0.1:60624 [rcache 0]
2015/04/09 14:17:43 skydns: ready for queries on local. for udp://127.0.0.1:60624 [rcache 0]
I0409 14:17:43.270362    2593 net.go:39] Got error &net.OpError{Op:"dial", Net:"tcp", Addr:(*net.TCPAddr)(0xc20884b950), Err:0x3d}, trying again: "127.0.0.1:60624"
I0409 14:17:43.374807    2593 master.go:620] OpenShift DNS listening at 127.0.0.1:60624
I0409 14:17:44.317131    2593 empties.go:28] AuthenticationError: Get https://127.0.0.1:60622: dial tcp 127.0.0.1:60622: connection refused
--- FAIL: TestOAuthBasicAuthPassword (4.87s)
    oauth_basicauth_test.go:223: unexpected error: listen tcp 127.0.0.1:60622: bind: address already in use
    oauth_basicauth_test.go:271: Unexpected error: an error on the server has prevented the request from succeeding
FAIL

We think it has something to do with unstarted HTTP listeners racing. Maybe.

[danmcp]

Any remaining issues here?

[smarterclayton]

Has been deflaked.

[smarterclayton]

Recurred in #2099

[liggitt]

not specific to OAuth test...

[smarterclayton]

.... The OS gives us the same address for two subsequent calls to listen. Added a retry loop.