We need to check that deploying process already was started if we execute the same deploy command through CLI.
Test Scenario:
1. Create new environment
2. Choose , Neutron, Vlan
3. Add 3 controllers
4. Provision 3 controllers
(fuel node --node-id x,x,x --provision --env x)
5. Start netconfig on second controller
(fuel node --node 2 --end netconfig --env x)
6. Start the deploy
(fuel node --node 2 --deploy --env-id x)
7. Start deploy again with the same parameters
(fuel node --node 2 --deploy --env-id x)
8. Execute fuel task command
Expected result:
If we execute the second command for deploying nodes, that are already in deploying status we must get Error, not new task creation.
Actual result:
http://paste.openstack.org/show/485141/ - for each command was created new task
VERSION:
feature_groups:
- mirantis
production: "docker"
release: "8.0"
api: "1.0"
build_number: "478"
build_id: "478"
fuel-nailgun_sha: "ae949905142507f2cb446071783731468f34a572"
python-fuelclient_sha: "4f234669cfe88a9406f4e438b1e1f74f1ef484a5"
fuel-agent_sha: "481ed135de2cb5060cac3795428625befdd1d814"
fuel-nailgun-agent_sha: "b2bb466fd5bd92da614cdbd819d6999c510ebfb1"
astute_sha: "b81577a5b7857c4be8748492bae1dec2fa89b446"
fuel-library_sha: "420c6fa5f8cb51f3322d95113f783967bde9836e"
fuel-ostf_sha: "ab5fd151fc6c1aa0b35bc2023631b1f4836ecd61"
fuel-mirror_sha: "b62f3cce5321fd570c6589bc2684eab994c3f3f2"
fuelmenu_sha: "fac143f4dfa75785758e72afbdc029693e94ff2b"
shotgun_sha: "63645dea384a37dde5c01d4f8905566978e5d906"
network-checker_sha: "9f0ba4577915ce1e77f5dc9c639a5ef66ca45896"
fuel-upgrade_sha: "616a7490ec7199f69759e97e42f9b97dfc87e85b"
fuelmain_sha: "6c6b088a3d52dd0eaf43d59f3a3a149c93a07e7e"
Nailgun process' separate provisioning/
As you may see there is no any check for already created deployment task in the manager and not without good reasons - currently we cannot by any way relate to task via nodes and vice versa as there is simply no connection between these two entities (please, note that the information could be acquired after all - via cluster object link; but it is not suitable for the case on the table as we need to operate on task in context of nodes not the whole cluster).
With that being said the only way to know whether deployment task is being performed on particular node is to check status of this node in nailgun db. But I consider this way as hacky and unreliable as this data may not fully correlates with reality.
AFAIK the link will be introduced in the scope of this blueprint [2]. Hence I propose to use it then in creating checks for task managers responsible for managing of actions for separate nodes.
[1]:https:/
[2]: https:/