cluster: Only pass a join address when in the process of joining a cluster

[aaronlehmann]

This code currently passes a random manager address when creating a new Node. This doesn't really make sense - we should only pass a join address on the initial join, or when retrying that join. An upcoming change to swarmkit will pay attention to JoinAddr when a node is already part of a cluster, so passing in the random value needs to be avoided.

Fixes #32980

cc @tonistiigi

[tonistiigi]

LGTM

[mlaventure]

LGTM

But I'd like for the test to be re-run given this is an old PR. I don't seem to be able to do so though. (cc @thaJeztah @tiborvass any idea why not?)

[thaJeztah]

No, no idea why it doesn't work, I think I've seen the same on some PR's. Let me try through Slack

[thaJeztah]

Yup, works through slack. Odd

[mlaventure]

All 3 Janky errors seem unrelated:

15:44:58 ----------------------------------------------------------------------
15:44:58 FAIL: docker_cli_exec_test.go:226: DockerSuite.TestExecStopNotHanging
15:44:58 
15:44:58 docker_cli_exec_test.go:248:
15:44:58     c.Fatal("Container stop timed out")
15:44:58 ... Error: Container stop timed out
15:44:58 
15:44:59 
15:44:59 ----------------------------------------------------------------------

15:57:41 ----------------------------------------------------------------------
15:57:41 FAIL: docker_cli_start_test.go:94: DockerSuite.TestStartPausedContainer
15:57:41 
15:57:41 docker_cli_start_test.go:98:
15:57:41     runSleepingContainer(c, "-d", "--name", "testing")
15:57:41 /go/src/github.com/docker/docker/pkg/testutil/cmd/command.go:64:
15:57:41     t.Fatalf("at %s:%d - %s", filepath.Base(file), line, err.Error())
15:57:41 ... Error: at cli.go:48 - 
15:57:41 Command:  /usr/local/bin/docker run -d -d --name testing busybox top
15:57:41 ExitCode: 125
15:57:41 Error:    exit status 125
15:57:41 Stdout:   cdb603cce5eb040a3003afb27fdb892b359a1cf3e39f2f68301c34ab7f2a1842
15:57:41 
15:57:41 Stderr:   /usr/local/bin/docker: Error response from daemon: endpoint with name testing already exists in network bridge.
15:57:41 
15:57:41 
15:57:41 Failures:
15:57:41 ExitCode was 125 expected 0
15:57:41 Expected no error
15:57:41 
15:57:41 
15:57:43 
15:57:43 ----------------------------------------------------------------------

17:05:55 ----------------------------------------------------------------------
17:05:55 FAIL: docker_cli_external_volume_driver_unix_test.go:446: DockerExternalVolumeSuite.TestExternalVolumeDriverBindExternalVolume
17:05:55 
17:05:55 docker_cli_external_volume_driver_unix_test.go:448:
17:05:55     dockerCmd(c, "run", "-d", "--name", "testing", "-v", "foo:/bar", "busybox", "top")
17:05:55 /go/src/github.com/docker/docker/pkg/testutil/cmd/command.go:64:
17:05:55     t.Fatalf("at %s:%d - %s", filepath.Base(file), line, err.Error())
17:05:55 ... Error: at cli.go:48 - 
17:05:55 Command:  /usr/local/bin/docker run -d --name testing -v foo:/bar busybox top
17:05:55 ExitCode: 125
17:05:55 Error:    exit status 125
17:05:55 Stdout:   d906acd135e600278a4463e3b82d0039430be0652d6309752d6369e599f0397e
17:05:55 
17:05:55 Stderr:   /usr/local/bin/docker: Error response from daemon: endpoint with name testing already exists in network bridge.
17:05:55 
17:05:55 
17:05:55 Failures:
17:05:55 ExitCode was 125 expected 0
17:05:55 Expected no error
17:05:55 
17:05:55 
17:05:56 
17:05:56 ----------------------------------------------------------------------

going to restart it for good mesure.

[mlaventure]

Alright, it passed, merging.

The other are flaky test. the TestStartPausedContainer seems to be an network issue though. I'll open an issue for it.