kolla-ansible reconfigure runs task "Gathering OSD IDs" on non-storage nodes

[Chris Hoge]

kolla-ansible reconfigure runs task "Gathering OSD IDs" from /ansible/roles/ceph/tasks/start_osds.yml on non-storage nodes. This will result in failure, as no directory '/var/lib/ceph/osd/' exists on non-storage nodes.

Output from failed run:

TASK [ceph : Gathering OSD IDs] ************************************************
fatal: [node1]: FAILED! => {"failed": true, "msg": "ERROR! 'osds' is undefined"}

[Jeffrey Zhang (jeffrey4l)]

could u post your inventory file out?

[Ravinder Kumar (rhcayadav)]

This bug is not reproduced in latest pike release. I have checked with 4 nodes( deployment node + controller & network node + storage node with ceph(2 osd) & compute node + Non storage node & compute node ).

Result :
root@manager1:~# kolla-ansible reconfigure -i multinode
Reconfigure OpenStack service : ansible-playbook -i multinode -e @/etc/kolla/globals.yml -e @/etc/kolla/passwords.yml -e CONFIG_DIR=/etc/kolla -e action=reconfigure -e serial=0 /usr/local/share/kolla-ansible/ansible/site.yml

PLAY RECAP *******************************************************************************************************************************
manager2.example.com : ok=354 changed=4 unreachable=0 failed=0
manager3.example.com : ok=95 changed=0 unreachable=0 failed=0
manager4.example.com : ok=70 changed=0 unreachable=0 failed=0

FYI:
 manager1.example.com deployment node
 manager2.example.com controller & network node
 manager3.example.com storage node with ceph(2 osd) & compute node
 manager4.example.com Non storage node & compute node