Simple cluster can not redeploy with new ceph node

[Dmitry Tyzhnenko]

api: '1.0'
astute_sha: bc60b7d027ab244039f48c505ac52ab8eb0a990c
auth_required: true
build_id: 2014-08-27_10-46-57
build_number: '481'
feature_groups:
- mirantis
fuellib_sha: 6f478caa9111e428b3c053eef44754f59ce7a2c2
fuelmain_sha: 74a97d500bb2fe9528f99771ccc2ec657ae3f76e
nailgun_sha: 04e3f9d9ad3140cd63a9b5a1a302c03ebe64fd0a
ostf_sha: 4dcd99cc4bfa19f52d4b87ed321eb84ff03844da
production: docker
release: '5.1'

Steps to reproduce:

1. Create simple cluster environment
2. Add 1 node with controller role
3. Add 1 node with compute role
4. Start environment
5. Add 1 node with ceph role
6. Start environment

Expected result:

- cluster is successfully deployed

Actual:

- deployment failed with the following error in puppet logs:

http://paste.openstack.org/show/101712/

[Matthew Mosesohn (raytrac3r)]

1353497 is just about saving and migrating SSH keys. It can't be exactly related.

[Sergey Vasilenko (xenolog)]

No, Matthew!

here was deployment without ceph. This deployment was OK.
after this was added ceph node. and deployed.

But controller and cinder nodes was not redeployed.
And this (new) env has no ceph part on controller and has no obligatory ssh keys on nodes.

It's no fuel-library bug -- it's a orchestration bug.

[Matthew Mosesohn (raytrac3r)]

Ok. If we have no keys generated by astute, then astute needs to for this case. I'll pass to fuel-python to see if we should adjust deployment serializer for this scenario. It's a very edge case that isn't normal.

Also this is not a high priority bug because there is a workaround: simply deploy ceph with initial deployment rather than adding it after deploying the controller.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/132636