Ansible: how handle servers on the same host? - ansible

Question

I have been trying for days to setup Ansible to use it for getting up a dev environment for my project, and secondly deploy to beta and live servers. The project is not that big but it seems that Ansible is not flexible enough when it comes to small projects.
Inventory
[development]
web_server ansible_connection=docker
db_server ansible_connection=docker
[production]
web_server ansible_host=10.10.10.10 ansible_user=tom ansible_connection=ssh
db_server ansible_host=10.10.10.10 ansible_user=tom ansible_connection=ssh
I want to keep the web_server and db_server aliases intact so I can switch between development and production in my scripts without hustle. The main issue is that I can't figure out how to create a playbook that will work nicely with the above setup.
This solution doesn't work since it runs all tasks twice!
---
- hosts: staging
tasks:
- name: Setup web server
command: uptime
delegate_to: web_server
- name: Setup db server
command: ls
delegate_to: db_server
This solution solves the above problem but it prints the wrong alias (web_server even when running the db task)!
---
- hosts: staging
run_once: true
tasks:
- name: Setup web servers
command: uptime
delegate_to: web_server
- name: Setup db servers
command: ls
delegate_to: db_server
This solution would be plausible but Ansible does not support access to an individual host from a group:
---
- hosts: staging:web_server
tasks:
- name: Deploy to web server
command: uptime
---
- hosts: staging:db_server
tasks:
- name: Deploy to db server
command: ls
Is there a way to accomplish what I want? Ansible feels quite restrictive until this point which is a bummer after all the praise I heard about it.
-------------------------- Edit after udondan's suggestion ----------------------
I tried udondan's suggestion and it seemed to work. However when I add a new group to the inventory it breaks.
[development]
web_server ansible_connection=docker
db_server ansible_connection=docker
[staging]
web_server ansible_host=20.20.20.20 ansible_user=tom ansible_connection=ssh
db_server ansible_host=20.20.20.20 ansible_user=tom ansible_connection=ssh
[production]
web_server ansible_host=10.10.10.10 ansible_user=tom ansible_connection=ssh
db_server ansible_host=10.10.10.10 ansible_user=tom ansible_connection=ssh
In this case the IP of the staging server (20.20.20.20) will be used when running the production playbook.

Answer 1

This solution doesn't work since it runs all tasks twice!
Assuming that hosts: staging is what you have defined in development this is the expected behavior. You defined a group of hosts and by running tasks or roles against this group all hosts of that group will be processed. By delegating the task to a different host you only force the task to be executed elsewhere, but the task still is executed for each host of the group.
I think what you want is this:
---
- hosts: web_server
tasks:
- name: Setup web server
command: uptime
- hosts: db_server
tasks:
- name: Setup db server
command: ls
Update after response:
The problem is you use the same hostnames for all environments and try to delegate them to different connectors. What Ansible though is doing is this:
It reads your inventory from top to bottom, finds the groups and processes the groups in alphabetical order. (development, production, staging) It finds the host web_server in group development, so it creates that group, adds that host and sets the var ansible_connection for this host. It proceed to the group production, again finds the host web_server so it adds it to the group production, and sets the vars ansible_host, ansible_user and ansible_connection. But this is not specific for this group. It is set for the hostname web_server, overriding the previous value of ansible_connection. Ansible continues to the staging group and again override all the settings. The host web_server belongs to all 3 groups and its vars have been merged, even if you have targeted only one group, e.g. ansible-playbook ... --limit=development. The limit only limits the play to matching hosts, but the host still belongs to all groups and has all the (merged) vars which have been defined.
It is highly recommended and best practice to have an inventory file per environment. Not specifically for the problem you are facing, but simply to add another layer of security to not accidentally do anything to your production hosts when you actually want to target staging boxes. But it will also solve your problem. So instead of one inventory file like you'd have:
inventory/development:
[development]
web_server
db_server
[development:vars]
ansible_connection=docker
inventory/staging:
[staging]
web_server
db_server
[staging:vars]
ansible_host=20.20.20.20
ansible_user=tom
ansible_connection=ssh
inventory/production:
[production]
web_server
db_server
[production:vars]
ansible_host=10.10.10.10
ansible_user=tom
ansible_connection=ssh
Then you'd call Ansible with the appropriate inventory file:
ansible-playbook -i inventory/staging ... playbook.yml
If you somehow can not do that and need a single inventory file you need to make sure the host names are unique so the connection details are not overridden.