Stop using ephemeral ports for integration tests

[liggitt]

We shouldn't use ephemeral ports for integration tests, because a free port can be taken by an outgoing request before we get a chance to start a listener on the port.

Also, we should indicate the range we want a port from so we can make sure the same port doesn't get chosen for master and dns, etc.

Fixes #1674

[smarterclayton]

Fixes #1674

[liggitt]

Hrmm... not the only issue, apparently:

masterAddr: "127.0.0.1:60666"
dnsAddr: "127.0.0.1:60568"
...
I0517 03:15:44.012282   13419 net.go:39] Got error &net.OpError{Op:"dial", Net:"tcp", Addr:(*net.TCPAddr)(0xc2093f1110), Err:0x6f}, trying again: "127.0.0.1:60666"
F0517 03:15:44.029771   13419 master.go:516] listen tcp 127.0.0.1:60666: bind: address already in use

[smarterclayton]

Awesome. Probably uses that port for free outgoing or something.

On May 17, 2015, at 12:08 AM, Jordan Liggitt notifications@github.com wrote:

Hrmm... not the only issue, apparently:

masterAddr: "127.0.0.1:60666"
dnsAddr: "127.0.0.1:60568"
...
I0517 03:15:44.012282 13419 net.go:39] Got error &net.OpError{Op:"dial", Net:"tcp", Addr:(*net.TCPAddr)(0xc2093f1110), Err:0x6f}, trying again: "127.0.0.1:60666"
F0517 03:15:44.029771 13419 master.go:516] listen tcp 127.0.0.1:60666: bind: address already in use
—
Reply to this email directly or view it on GitHub.

[smarterclayton]

two options:

1. Pass a listener in
2. Don't rely on the OS to pick a free port like this, instead search ourselves in a lower range

[liggitt]

I'd rather search ourselves in a lower range. Will update tomorrow

[liggitt]

[test]

[openshift-bot]

continuous-integration/openshift-jenkins/test SUCCESS (https://ci.openshift.redhat.com/jenkins/job/test_pull_requests_origin/2245/)

[smarterclayton]

LGTM [merge]

[openshift-bot]

continuous-integration/openshift-jenkins/merge SUCCESS (https://ci.openshift.redhat.com/jenkins/job/merge_pull_requests_origin/1988/) (Image: devenv-fedora_1562)

[openshift-bot]

Evaluated for origin up to 54a895b