$$network_metadata["vips"]["public"] is :undef on 'virt' role

[Alexandr Kostrikov]

Detailed bug description:

During deploy of reduced_footprint tasks we have an error on 'virt' role:
2016-04-05 23:58:49 +0000 Puppet (err): $$network_metadata["vips"]["public"] is :undef, not a hash or array at /etc/puppet/modules/osnailyfacter/manifests/globals/globals.pp:233 on node node-1.test.domain.local

Full stacktrace is here [0].

Steps to reproduce:
Run deployment of reduced footprint with virtual role.
Can be done with fuel-qa with group 'spawn_one_vm_on_one_virt_node'

Expected results:
Deployment is done on lack of hardware.

Actual result:
Deployment has failed on [gloabal] task

Additional information:
That is reproduced on swarm and environment can be found at[1]
That also is a swram blocker due to fact that it failed 3 jobs:
 - Spawn one vm node on one slave node
 - Spawn two vm nodes on one slave node
 - Spawn three vm nodes across three slave nodes

[0] http://paste.openstack.org/show/493138/
[1] https://product-ci.infra.mirantis.net/job/9.0.system_test.ubuntu.reduced_footprint/

[Bartosz Kupidura (zynzel)]

Looks like problem with nailgun. vips key in astute (node-1/etc/hiera/astute.yaml) is empty:
network_metadata:
  nodes:
    node-1:
      fqdn: node-1.test.domain.local
      name: node-1
      network_roles: {admin/pxe: 10.109.11.3, aodh/api: 10.109.12.2, ceilometer/api: 10.109.12.2,
        ceph/public: 10.109.13.2, ceph/radosgw: 10.109.14.2, ceph/replication: 10.109.13.2,
        cinder/api: 10.109.12.2, cinder/iscsi: 10.109.13.2, ex: 10.109.14.2, fw-admin: 10.109.11.3,
        glance/api: 10.109.12.2, glance/glare: 10.109.12.2, heat/api: 10.109.12.2,
        horizon: 10.109.12.2, ironic/api: 10.109.12.2, keystone/api: 10.109.12.2,
        management: 10.109.12.2, mgmt/corosync: 10.109.12.2, mgmt/database: 10.109.12.2,
        mgmt/memcache: 10.109.12.2, mgmt/messaging: 10.109.12.2, mgmt/vip: 10.109.12.2,
        mongo/db: 10.109.12.2, murano/api: 10.109.12.2, murano/cfapi: 10.109.12.2,
        neutron/api: 10.109.12.2, neutron/floating: null, neutron/mesh: 10.109.15.1,
        neutron/private: null, nova/api: 10.109.12.2, nova/migration: 10.109.12.2,
        public/vip: 10.109.14.2, sahara/api: 10.109.12.2, storage: 10.109.13.2, swift/api: 10.109.12.2,
        swift/replication: 10.109.13.2}
      node_roles: [compute, virt]
      nova_cpu_pinning_enabled: false
      nova_hugepages_enabled: false
      swift_zone: '1'
      uid: '1'
      user_node_name: slave-01_compute_virt
  vips: {}

[Aleksey Kasatkin (alekseyk-ru)]

No default VIPs should be serialized for environment that does not have controllers. All default VIPs (management/vrouter, management/haproxy, public/vrouter, public/haproxy) have node_roles=['controller', 'primary-controller']. This environment has just one node with ['compute', 'virt'] roles. So, no VIPs should be there.

[Aleksey Kasatkin (alekseyk-ru)]

Also, I see just two successful runs of this system test (March, 9 and 10). Other 30 has failed.

I don't see any changes in fuel-web on Mar, 10-11 that can be related to this (to VIPs, node roles, etc.).

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/302793

[OpenStack Infra (hudson-openstack)]

Related fix proposed to branch: master
Review: https://review.openstack.org/302965

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/302793
Committed: https://git.openstack.org/cgit/openstack/fuel-library/commit/?id=1ddd22a11415cfb005e061007dee618cb848ae34
Submitter: Jenkins
Branch: master

commit 1ddd22a11415cfb005e061007dee618cb848ae34
Author: Michael Polenchuk <email address hidden>
Date: Thu Apr 7 17:06:07 2016 +0300

    Dig vips out from network metadata

    In order to prevent the refer to an empty vips for virt/compute roles
    (there are no controllers in cluster yet) fallback to default values.

    Change-Id: Ic6dfebd2b032f59312b0f0af60d99db6d28763d7
    Closes-Bug: #1566796

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/mitaka
Review: https://review.openstack.org/305011

[Alexandr Kostrikov (akostrikov-mirantis)]

Fix has not yet in iso on swarm:
https://product-ci.infra.mirantis.net/job/9.0.system_test.ubuntu.reduced_footprint/76/console

fuel-library9.0-9.0.0-1.mos8288.noarch
* Tue Apr 12 2016 Jenkins <email address hidden> - 9.0.0-1.mos8288
- 63b6957 Merge "Switch from RUN: to ROLE: annotation" into stable/mitaka
- 5b1192e Implement Glance Glare service

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/305011
Committed: https://git.openstack.org/cgit/openstack/fuel-library/commit/?id=e544a5102cbbf8fb20712fa0ff79c891a7e1f595
Submitter: Jenkins
Branch: stable/mitaka

commit e544a5102cbbf8fb20712fa0ff79c891a7e1f595
Author: Michael Polenchuk <email address hidden>
Date: Thu Apr 7 17:06:07 2016 +0300

    Dig vips out from network metadata

    In order to prevent the refer to an empty vips for virt/compute roles
    (there are no controllers in cluster yet) fallback to default values.

    Change-Id: Ic6dfebd2b032f59312b0f0af60d99db6d28763d7
    Closes-Bug: #1566796
    (cherry picked from commit 1ddd22a11415cfb005e061007dee618cb848ae34)

[Alexandr Kostrikov (akostrikov-mirantis)]

Checked on iso: fuel-9.0-201-2016-04-14_14-00-00.iso