[nailgun] Task 'deploy' remains in DB even if deployment was failed to start

2014-08-08 07:55:46.894 DEBUG [7f09f7f53740] (logger) Request PUT /api/clusters/3/changes from 10.20.0.1:38481 {}

2014-08-08 07:55:46.920 DEBUG [7f09f7f53740] (logger) Response code '400 Bad Request' for PUT /api/clusters/3/changes from 10.20.0.1:38481

2014-08-08 08:43:12 DEBUG [7f09f7f53740] (logger) Response code '400 Bad Request' for PUT /api/clusters/3/stop_deployment from 10.20.0.1:40025

[root@fuel tmp]# fuel env
id | status | name | mode | release_id | changes | pending_release_id
---|-------------|--------|-----------|------------|--------------------------------------|-------------------
3 | operational | 3e4rty | multinode | 2 | [{u'node_id': 6, u'name': u'disks'}] | None

Reproduced on :

  production: "docker"
  release: "5.1"
  api: "1.0"
  build_number: "414"
  build_id: "2014-08-08_11-23-51"
  astute_sha: "b52910642d6de941444901b0f20e95ebbcb2b2e9"
  fuellib_sha: "d699fc178559e98cfd7d53b58478b46553ffe39e"
  ostf_sha: "e33390c275e225d648b36997460dc29b1a3c20ae"
  nailgun_sha: "5bc33457e5a1f108b071ed0ef2a771ea0b610b22"
  fuelmain_sha: "16c54168143061724e635a20a99545a756725b49"

There is an backend bug:
The main problem is that backend starts task even when wrong changes are coming.

It is easy reproducible on fakeUI:
Deploy cluster - you'll see the cluster with disabled "Deploy changes" button. Then you can to try to put any changes to cluster using browser console, or just edit "Deploy changes" button using "inspect element" - remove the "disabled" class, then you can click on this button and start deployment with no changes on the cluster actually.
The result:
PUT http://0.0.0.0:8000/api/clusters/1/changes
400 Bad Request
Response: No changes to deploy
 - so there is an expected behavior.

But! For any case - backend starts an mysterious task.
It is easy visible in browser console:
[{"status": "running", "name": "deploy", "cluster": 1, "result": {}, "progress": 0, "message": null, "id": 27, "uuid": "a078732f-d376-4b7c-838b-41664a85a4b9"}]
UI also are viewing this task - so "progress bar" and "stop deployment" elements are showed for the cluster.

But - backend doesn't see it - So when we trying to stop this task: we'll se the " Nothing to stop - deployment is not running on environment '1'"

Fix proposed to branch: master
Review: https://review.openstack.org/114568

Reviewed: https://review.openstack.org/114568
Committed: https://git.openstack.org/cgit/stackforge/fuel-web/commit/?id=bc9e377dbe010732bc2ba47161ed9d433998e07b
Submitter: Jenkins
Branch: master

commit bc9e377dbe010732bc2ba47161ed9d433998e07b
Author: Aleksey Kasatkin <email address hidden>
Date: Fri Aug 15 18:15:54 2014 +0300

    Delete task 'deploy' before raise exception on deployment start

    Closes-Bug: #1354401

    Change-Id: I5906cbfc094e0d0ca67158ba9fbec233eefb014e

i'm still meeting this problem in fuel 6.1 when i add node to a present environment。 probably because i stopped the deployment several times.
Can anyone tell me how to fixed it temporarily? I don't want to redeploy the environment because it's in use.