Hi all.
I have a master deployment playbook that more or less includes all of my servers, which I use with tags to deploy only the apps/configs I want at the time. Hosts are grouped into hostgroups by roles and each hostgroup has it’s own pre/post tasks for handling load-balancers, etc. It looks something like this:
-
hosts: tag_role_hostgroup1
serial: 1
pre_tasks: -
name: Remove from lb
…
tags: pre_tasks
roles: -
{ role: app1, tags: app1 }
-
{ role: app2, tags: app2 }
post_tasks: -
name: Register to lb
…
tags: post_tasks -
hosts: tag_role_hostgroup2
serial: 1
pre_tasks:
- name: Some preps
…
tags: pre_tasks
roles: - { role: app2, tags: app2 }
- { role: app3, tags: app3 }
post_tasks: - name: Cleanup
…
tags: post_tasks
So far so good. It works pretty well as I can think about applications when deploying, and each hostgroup will take care of itself regarding load-balancers and all. Problem is when deploying for example app1 or app3. In this case one play will have no roles selected, but since I’m including “pre_tasks,post_tasks” as tags its pre/post tasks will still run, bringing machines down and up again with no work done.
Do you have an idea of how to tackle this? I prefer to not create a playbook per app as it’s a lot less powerful and flexible. I was thinking of making ansible run pre/post tasks only if some task in roles/tasks is going to run, but that’s kind of a last resort.
Thanks for Ansible, btw!