]> git.apps.os.sepia.ceph.com Git - ceph-ansible.git/commit
take-over: fix bug when trying to override variable
authorGuillaume Abrioux <gabrioux@redhat.com>
Thu, 17 May 2018 15:29:20 +0000 (17:29 +0200)
committerSébastien Han <seb@redhat.com>
Fri, 18 May 2018 09:21:37 +0000 (11:21 +0200)
commit3c2f33221ccafbc4f8d0151e344edbe5cc728635
treeef87151380a4b45fa86ea5b75ea2dc8579783f25
parent6321a651d1de74c7079e7abb5daca75ed58a58c3
take-over: fix bug when trying to override variable

A customer has been facing an issue when trying to override
`monitor_interface` in inventory host file.
In his use case, all nodes had the same interface for
`monitor_interface` name except one. Therefore, they tried to override
this variable for that node in the inventory host file but the
take-over-existing-cluster playbook was failing when trying to generate
the new ceph.conf file because of undefined variable.

Typical error:

```
fatal: [srvcto103cnodep01]: FAILED! => {"failed": true, "msg": "'dict object' has no attribute u'ansible_bond0.15'"}
```

Including variables like this `include_vars: group_vars/all.yml` prevent
us from overriding anything in inventory host file because it
overwrites everything you would have defined in inventory.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1575915
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 415dc0a29b10b28cbd047fe28eb4dd38419ea5dc)
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
infrastructure-playbooks/take-over-existing-cluster.yml