uft-cluster-config hangs

[exarkun]

$ uft-flocker-config cluster.yml
Initialized cluster CA.
Created control cert.
Generated fc43dc40-8a58-41c8-a664-05405f1e073b for ....
Generated 4c763764-834b-42a8-917c-9bac4817c02d for ....
Generated f96c7b9f-b7e2-46d1-b66b-33f62c3e568e for ....
Created user key for coreuser
Making /etc/flocker directory on all nodes
Uploading keys to respective nodes:
 * Uploaded control cert & key to control node.
 * Uploaded cluster cert to ....
 * Uploaded agent.yml to ....
 * Uploaded node key to ....
 * Uploaded node key to ....
 * Uploaded node crt to ....
 * Uploaded node crt to ...
 * Uploaded cluster cert to ....
 * Uploaded agent.yml to ....
 * Uploaded node key to ....
 * Uploaded cluster cert to ....
 * Uploaded control key to ....
 * Uploaded control crt to ....
 * Uploaded cluster cert to ....
 * Uploaded agent.yml to ....
 * Uploaded node crt to ....

After the last line, many minutes pass with no further progress. The ssh process being run by the uft-cluster-config container is blocked on select().

[wallnerryan]

does this happen every time you run it? I haven't come across this issue, just curious, and ran it last night.

[exarkun]

If I try to run it again it fails with an error very quickly:

$ uft-flocker-config cluster.yml
Error: Unable to write certificate file. File exists /pwd/cluster.crt
main function encountered error
Traceback (most recent call last):
  File "/opt/flocker/bin/flocker-config", line 9, in <module>
    load_entry_point('UnofficialFlockerTools==0.5', 'console_scripts', 'flocker-config')()
  File "/opt/flocker/lib/python2.7/site-packages/unofficial_flocker_tools/config.py", line 147, in _main
    react(main, sys.argv[1:])
  File "/opt/flocker/lib/python2.7/site-packages/twisted/internet/task.py", line 875, in react
    finished = main(_reactor, *argv)
  File "/opt/flocker/lib/python2.7/site-packages/twisted/internet/defer.py", line 1253, in unwindGenerator
    return _inlineCallbacks(None, gen, Deferred())
--- <exception caught here> ---
  File "/opt/flocker/lib/python2.7/site-packages/twisted/internet/defer.py", line 1107, in _inlineCallbacks
    result = g.send(result)
  File "/opt/flocker/lib/python2.7/site-packages/unofficial_flocker_tools/config.py", line 21, in main
    c.run("flocker-ca initialize %s" % (c.config["cluster_name"],))
  File "/opt/flocker/lib/python2.7/site-packages/unofficial_flocker_tools/utils.py", line 208, in run
    result = subprocess.check_output(command, shell=True)
  File "/usr/lib/python2.7/subprocess.py", line 573, in check_output
    raise CalledProcessError(retcode, cmd, output=output)
subprocess.CalledProcessError: Command 'flocker-ca initialize jp_coreos_flocker_testing' returned non-zero exit status 1

Perhaps I'll move all the generated authentication files aside and see if that lets me run it again...

[exarkun]

If I do that then I get a different error:

Retrying running ['-o', 'LogLevel=error', '-o', 'UserKnownHostsFile=/dev/null', '-o', 'StrictHostKeyChecking=no', '-i', '/host/tmp/flocker-coreos/jean-paulcalderoneinsecure-temporary.pem', 'root@....', "bash -c 'echo; echo\necho > /tmp/flocker-command-log\ndocker run --restart=always -d --net=host --privileged \\\n    -v /etc/flocker:/etc/flocker \\\n    -v /var/run/docker.sock:/var/run/docker.sock \\\n    --name=flocker-container-agent \\\n    clusterhq/flocker-container-agent\ndocker run --restart=always -d --net=host --privileged \\\n    -e DEBUG=1 \\\n    -v /tmp/flocker-command-log:/tmp/flocker-command-log \\\n    -v /flocker:/flocker -v /:/host -v /etc/flocker:/etc/flocker \\\n    -v /dev:/dev \\\n    --name=flocker-dataset-agent \\\n    clusterhq/flocker-dataset-agent\n'"] on .... given result 'Process exited with error code 1: \n\nError response from daemon: Conflict. The name "flocker-container-agent" is already in use by container e296946b20e2. You have to delete (or rename) that container to be able to reuse that name.\nError response from daemon: Conflict. The name "flocker-dataset-agent" is already in use by container 09b2ba7ae2e4. You have to delete (or rename) that container to be able to reuse that name.\n'...

[wallnerryan]

Yeah this is an issue because it doesn't clean up certs on re-run. +1 to making it re-runnable. Relates to or may duplicate #42

[exarkun]

Nevertheless, uft-flocker-volumes now succeeds... Is the config command just expected to take a long time and not generate any progress updates?

[wallnerryan]

yeah, it can potentially take a little bit of time. I've seen it hang before but not for too long (<1 to 2 min). Would be nice to have progress bars or updates.

[lukemarsden]

I've seen it hang like this before. Maybe if we try and scp at just the wrong time, TCP screws us. We could add timeouts on each network operation.

[lukemarsden]

Also, +1 for fixing #42