swarm bay is "CREATE_COMPLETE" but the bay is not usable acutally

[Eli Qiao]

I created a bay with atomic-5 image with coe=swarm

+--------------------+------------------------------------------+
| Property | Value |
+--------------------+------------------------------------------+
| status | CREATE_COMPLETE |
| uuid | 0c7fea29-cb66-4d6a-bcd7-698f7e6243b9 |
| status_reason | Stack CREATE completed successfully |
| created_at | 2015-09-28T02:34:54+00:00 |
| updated_at | 2015-09-28T02:37:11+00:00 |
| bay_create_timeout | 0 |
| api_address | 172.24.4.206 |
| baymodel_id | 57614820-1871-45a6-af3a-024aac27c298 |
| node_count | 1 |
| node_addresses | [u'172.24.4.207'] |
| master_count | 1 |
| discovery_url | token://c0717094a987deeb88d6e7b5aba6d14a |
| name | swarmbay |

after some time I see it is 'CREATE_COMPLETE' state, but when I login into the bay's vms
swarm-master, and swarm-node failed to run..
[fedora@swarmbay-pn4pjq4gg4zw-swarm-master-bfx7doenlits ~]$ sudo docker ps -a
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
6aaadf3a26d4 swarm:0.2.0 "/swarm manage -H tc 42 minutes ago Exited (137) 42 minutes ago swarm-manager
417356a72d82 swarm:0.2.0 "/swarm join --addr 53 minutes ago Exited (2) 53 minutes ago swarm-agent

we need some validation before the bay is usable.

[hongbin (hongbin034)]

Basically, we need to fix the swarm template to make it work with the new image. This is a high priority bug.

[Ton Ngo (ton-i)]

@hongbin: I think you mean this bug, which documents the specific problem with "docker run" in swarm cluster:

https://bugs.launchpad.net/magnum/+bug/1499607?

The bug 1500291 here is intended for the more general scenario, where the cluster may complete as far as Heat is concerned, but the services in the cluster are not running correctly. This can happen with k8s or swarm cluster. Basically we need to add some more validation code to check on the health of the services in the cluster.

[Egor Guz (eghobo)]

@hongbin: I am not sure it's Swarm template issue. I think something wrong with image (or we configure it wrong), because Kub instances has the same symptoms .

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/228762

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/228762
Committed: https://git.openstack.org/cgit/openstack/magnum/commit/?id=832aa604363aebef178129a84d05a6b71de48cae
Submitter: Jenkins
Branch: master

commit 832aa604363aebef178129a84d05a6b71de48cae
Author: Eli Qiao <email address hidden>
Date: Tue Sep 29 14:58:04 2015 +0800

    Swarm: Set to CREATE_FAILED status if swarm services not started

    This patch checks $NODE_SERVICES status before send SUCCESS single to heat cfn,
    this will make sure the bay is usable.

    Change-Id: Ie232c578c5c27b1842965bdda481096fb0b5c820
    Closes-Bug: #1500291