This project is meant to store ansible roles for managing the nodes in the ceph testing labs.
As this repo only contains roles, it does not define the ansible inventory or
any associated group_vars or host_vars. However, it does depend on these
things existing in a separate repository or otherwise accesible by these roles
when they are used. Any vars a role needs should be added to its
defaults/main.yml
file to document what must be defined per node or group
in your inventory.
This separation is important because we have multiple labs we manage with these
same roles and each lab has different configuration needs. We call these our
secrets
or *-secrets
repos throughout the rest of the documention and
in the roles.
Besides the inventory, secrets
repos also may contain certain secret or
encrypted files that we can not include in ceph-cm-ansible for various reasons.
The directory structure for one of our secrets
repos is:
├── ansible ├── inventory │ ├── group_vars │ │ ├── all.yml │ │ ├── cobbler.yml │ │ ├── testnodes.yml │ │ ├── teuthology.yml │ │ └── typica.yml │ └── sepia └── secrets └── entitlements.yml
Refer to Step 2 below for instructions on how to setup a secrets
repo for
use by ceph-cm-ansible. If set up this way, -i is not necessary for
ansible-playbook to find the repo. However, you can choose your own setup and
point to the secrets
repo with -i if you prefer.
NOTE: Some playbooks require specific groups to be defined in your
inventory. Please refer to hosts
in the playbook you want to use to ensure
you've got the proper groups defined.
All variables should be defined in defaults/main.yml
for the role they're
primarily used in. If the variable you're adding can be used in multiple roles
define it in defaults/main.yml
for both roles. If the variable can contain
a reasonable default value that should work for all possible labs then define
that value in defaults/main.yml
as well. If not, you should still default
the variable to something, but make the tasks that use the variable either fail
gracefully without that var or prompt the user to define it if it's mandatory.
If the variable is something that might need to be defined with a value
specific to the lab in use, then it'll need to be added to your secrets
repo as well. Variables in group_vars/all.yml
will apply to all nodes
unless a group_var file exists that is more specific for that node. For
example, if you define the var foo: bar
in all.yml
and the node you're
running ansible against exists in the testnodes
group and there is a
group_vars/testnodes.yml
file defined with foo: baz
included in it then
the role using the variable will use the value defined in testnodes.yml
.
The playbook you're using knows which group_var file to use because of the
hosts
value defined for it.
We assume that your SSH key is present and active for passwordless access to the "ubuntu" shell user on the hosts that ansible will manage.
You can use pip:
pip install ansible
or use the OS package manager:
yum install ansible
Clone the secrets repository and symlink the hosts
and secrets
directories into place:
cd $HOME/dev/ git clone git@..../ceph-ansible-secrets.git sudo mv /etc/ansible/hosts /etc/ansible/hosts.default sudo ln -s /path/to/ceph-ansible-secrets/ansible/inventory /etc/ansible/hosts sudo ln -s /path/to/ceph-ansible-secrets/ansible/secrets /etc/ansible/secrets
Clone the main Ceph ansible repository:
git clone git@..../ceph-cm-ansible.git cd ceph-cm-ansible
If you have any new hosts on which you'd like to run ansible, or if you're
using separate testing VMs, edit the files in /etc/ansible/hosts
to add
your new (or testing) hosts:
vi /etc/ansible/hosts/<labname>
If you don't need to test on any new hosts, you can skip this step and just use
/etc/ansible/hosts
as-is.
You can now run ansible-playbook
:
vi myplaybook.yml ansible-playbook myplaybook.yml -vv --check --diff
This will print a lot of debugging output to your console.
Ansible runs using the "cm" shell account.
Let's say you've created a new VM host using downburst. At this point you should have a new VM with the "ubuntu" UID present. The problem is that Ansible uses the "cm" user. In order to get that UID set up:
Add your host to the inventory. Look in your lab's
secrets
repository, in theansible/inventory/
directory, and add your new node.Run the
cephlab.yml
playbook, limited to your new host "mynewhost":ansible-playbook -vv --limit mynewhost cephlab.yml