'rabbitmq' task doesn't re-install the package 'rabbitmq-server'

[Dennis Dmitriev]

Steps to reproduce:
    1. Create a cluster with 3 controller,
    2. Deploy cluster,
    3. Uninstall the package 'rabbitmq-server' from all controllers,
    4. Start the task 'rabbitmq': fuel node --node 1,2,3 --tasks rabbitmq --env 1

Expected result: task finished successfully, pcs resource for rabbitmq are started and rabbit is working on all controllers.

Actual result: task 'rabbitmq' failed with errors in puppet.log on the primary controller:

Puppet (debug): Fail: Execution of '/usr/sbin/rabbitmqctl -q list_users' returned 2: Error: {aborted,{no_exists,[rabbit_user,{inte
rnal_user,'_','_','_'}]}}

On other controllers the package 'rabbitmq-server' even weren't installed.

Workaround: if the 'rabbitmq-server' is installed manually on all controllers, then the task 'rabbitmq' is passed succesfully.

It is blocker for https://bugs.launchpad.net/fuel/+bug/1398009

[Dennis Dmitriev (ddmitriev)]

api: '1.0'
astute_sha: b1f37a988e097175cbbd14338286017b46b584c3
auth_required: true
build_id: 2015-07-23_10-59-34
build_number: '82'
feature_groups:
- mirantis
fuel-agent_sha: bc25d3b728e823e6154bac0442f6b88747ac48e1
fuel-library_sha: 58d94955479aee4b09c2b658d90f57083e668ce4
fuel-ostf_sha: 94a483c8aba639be3b96616c1396ef290dcc00cd
fuelmain_sha: 68871248453b432ecca0cca5a43ef0aad6079c39
nailgun_sha: d1087923e45b0e6d946ce48cb05a71733e1ac113
openstack_version: 2015.1.0-7.0
production: docker
python-fuelclient_sha: 471948c26a8c45c091c5593e54e6727405136eca
release: '7.0'
release_versions:
  2015.1.0-7.0:
    VERSION:
      api: '1.0'
      astute_sha: b1f37a988e097175cbbd14338286017b46b584c3
      build_id: 2015-07-23_10-59-34
      build_number: '82'
      feature_groups:
      - mirantis
      fuel-agent_sha: bc25d3b728e823e6154bac0442f6b88747ac48e1
      fuel-library_sha: 58d94955479aee4b09c2b658d90f57083e668ce4
      fuel-ostf_sha: 94a483c8aba639be3b96616c1396ef290dcc00cd
      fuelmain_sha: 68871248453b432ecca0cca5a43ef0aad6079c39
      nailgun_sha: d1087923e45b0e6d946ce48cb05a71733e1ac113
      openstack_version: 2015.1.0-7.0
      production: docker
      python-fuelclient_sha: 471948c26a8c45c091c5593e54e6727405136eca
      release: '7.0'

[Vladimir Kuklin (vkuklin)]

This is a good but synthetic use case - noone in sane mind will uninstall primary software from the controller nodes. I am this lowering the priority of this bug.

[Alex Schultz (alex-schultz)]

I took a look at this and I think it's actually a bug in the upstream rabbitmq module we use. Just throwing it out there as something that should probably be communicated upstream.

[Tomasz Adam Jaroszewski (tjaroszewski)]

I've checked this issue in lab environment.
Everything works are described but one thing is missing - removal/cleaning of rabbit resource after uninstall of rabbitmq-server:
crm configure delete p_rabbitmq-server && crm resource cleanup p_rabbitmq-server

With these step taken tasks is successful: pcs "rabbit" resource is created and rabbit-server "reinstallation" is running correct.

There is also a additional workaround (if removal of resource is not acceptable)- restart of rabbitmq-server on master node during check:
2015-07-29 12:58:42 +0000 Puppet (debug): Fail: Execution of '/usr/sbin/rabbitmqctl -q list_users' returned 2: Error: {aborted,{no_exists,[rabbit_user,{internal_user,'_','_','_'}]}};

[Matthew Mosesohn (raytrac3r)]

As far as I can tell, this isn't a bug. Only an admin could manually do this