Hi,
We are creating playbooks for our servers that include creating instances with rackspace. We have problems with setting up the different inventory files with the appropriate vars. To illustrate the problem I created a small test case:
DIR
./a.yml
./b.yml
./test
a.yml contains:
The host is a member of two groups.
In this case, rightfully, variables are pulled from both.
This is by design – you may select hosts based on purpose (such as webservers), while assigning other variables based on geography.
Thanks for replying and clarifying.
So then my question is this: Both application A and B require a local connection and variables to communicate with rackspace (like hostname for instance). Do you have some pointers how I should setup my inventory file and were to put my variables? The playbook for both apps is basically:
Like the example below one of the apps fails because it (rightfully) resolves to the wrong hostname.
Thanks again.
Lars
I’m having difficulty between drawing the lines between “command echo” and your cloud integration.
The rax module has a parameter “instance-name” and will create that instance with that name if it does not exist.
http://ansibleworks.com/docs/modules.html#rax
You can also use the “exact_count” parameter to request an exact number of instances starting with a given name, and it will slay extra instances if you have too many.
BTW, you don’t pre-create hosts when using cloud machines.
You should be using the rackspace dynamic inventory script.
You would still use add_host, but if you are trying to predefine the system in your INI based inventory, that’s not really the way it’s meant to work.
The cloud is meant to be authorative.
Can you still use group variables? Sure. Should you it to define hostnames prior to creating instances – no.
I am doing just that. Here is the code:
a.yml and b.yml start with the same rackspace setup. How can I not have hostname, or the flavor or the image as a variable?
hosts: local-rackspace
connection: local
remote_user: root
tasks:
name: Rackspace | Create Rackspace server instance
local_action:
module: rax
credentials: ~/.rax
name: “{{ hostname }}”
flavor: “{{ rackspace_flavor }}”
image: “{{ rackspace_image }}”
region: LON
state: present
wait: yes
wait_timeout: 900
files:
/root/.ssh/authorized_keys: ~/.ssh/id_rsa.pub
register: rax
name: Rackspace | Add new instance to host group
action: add_host hostname={{ hostname }} ansible_ssh_host=${item.accessIPv4} groupname={{ groupname }}
with_items: ${rax.instances}
My slimmed down inventory file:
[local-rackspace]
localhost ansible_python_interpreter=/usr/local/bin/python
[a]
[b]
[acceptance-a:children]
a
local-rackspace
[a:vars]
hostname=test.a.com
rackspace_flavor=4
groupname=a
[acceptance-b:children]
b
local-rackspace
[acceptance-b:vars]
hostname=test.b.com
rackspace_flavor=3
groupname=b
rackspace_image is defined in group_vars/all since we use the same base image for our apps.
Maybe my earlier example was to simplistic but I thought it was easier to explain my question since it doesn’t require a rackspace account to test it.
So my question is still: how can I make sure that I pass the correct variables to the rax module in this case? Or I just hardcode them in the file? I’d really like it to make it a role so I can use it for my other apps as well, but being unable to use variables makes it impossible. Now I am sure I am missing something really obvious here, but I can’t figure out what…
Thanks again for your help.
Lars
Ok I get how it supposed to work now. I’d like to share our setup.
We created a rackspace role and pass in the different variables (different apps have different flavors on different environments). We no longer add the local-rackspace group to the application group and add specific vars to the local-rackspace group on specific environments(acceptance/production). Setup is now like this:
playbook app_a:
./appa.yml