Ceph-rgw deployment problems

[wangwei]

This is my configuration in the globals.yml

```
enable_ceph: "yes"
enable_ceph_rgw: "no"
```

This is my configuration in the multinode-inventory

```
[storage-mon]
ceph01 ansible_user=irteam network_interface=em1 api_interface=em1
ceph02 ansible_user=irteam network_interface=em1 api_interface=em1

[storage-osd]
ceph01 ansible_user=irteam network_interface=em1 api_interface=em1
ceph02 ansible_user=irteam network_interface=em1 api_interface=em1
[storage]
dev01 ansible_user=root storage_interface=eth0 cluster_interface=eth0
dev02 ansible_user=root storage_interface=eth0 cluster_interface=eth0

[ceph-mon:children]
storage-mon

[ceph-rgw:children]
storage

[ceph-osd:children]
storage-osd
```

This is my deployment command:

```
kolla-ansible/tools/kolla-ansible deploy --configdir /xxx -i /xxx/multinode-inventory --passwords /xxx/passwords.yml --tags ceph -e openstack_release=prod-xxxxx
```

The deployment will produce an error, but the deployment process doesn't stop.

```
TASK [ceph : Copying over config.json files for services] *****************************************************************
failed: [dev02] (item=ceph-mon) => {"failed": true, "item": "ceph-mon", "msg": "AnsibleUndefinedVariable: 'dict object' has no attribute u'ansible_em1'"}
failed: [dev01] (item=ceph-mon) => {"failed": true, "item": "ceph-mon", "msg": "AnsibleUndefinedVariable: 'dict object' has no attribute u'ansible_em1'"}
failed: [dev02] (item=ceph-osd) => {"failed": true, "item": "ceph-osd", "msg": "AnsibleUndefinedVariable: 'dict object' has no attribute u'ansible_em1'"}
failed: [dev01] (item=ceph-osd) => {"failed": true, "item": "ceph-osd", "msg": "AnsibleUndefinedVariable: 'dict object' has no attribute u'ansible_em1'"}
```
The reason for the error is that the ceph-rgw node will generate the ceph-mon configuration file. In the ceph installation process, there is no distinction between mon, osd and rgw nodes, but in all nodes are generated mon, osd and rgw configuration file, so that the rgw node to generate a mon container configuration file(config.json) will report above error.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/496608

[Eduardo Gonzalez (egonzalez90)]

Hi, i cannot reproduce the issue? Could you try to see if still present? Maybe was fixed on other patch no attached to this bug id

[wangwei (wangwei-david)]

Hi Eduardo Gonzalez,

I tested the above problem and can solve it in this way:

ceph-node1 ansible_user=usera network_interface=eth0 api_interface=eth0 storage_interface=eth0 cluster_interface=eth0
ceph-node2 ansible_user=usera network_interface=em1 api_interface=em1 storage_interface=em1 cluster_interface=em1

After all four network cards are specified, there will be no conflicts.

However, there is still a problem with the current deployment. That is, if I only configure a single rgw node, there is only the ceph-rgw service, but other configuration folders will be generated during deployment, such as ceph-mon, ceph-osd and so on.