managing multi-node environments

Archives Ansible Project
1

I have environments that contain different nodes of different roles/types like: databases, webservers, load-balancers.
Sometimes I need to call every server of the same role/type, but sometimes I need to address an environment.

e.g.:

[databases]
test1.db.my.com ansible_ssh_host=192.168.1.200 ansible_ssh_user=…
test2.db.my.com ansible_ssh_host=192.168.1.201 ansible_ssh_user=…

[webservers]
test1.web.my.com ansible_ssh_host=192.168.1.300 ansible_ssh_user=…
test2.web.my.com ansible_ssh_host=192.168.1.301 ansible_ssh_user=…

[load-balancers]
test1.lb.my.com ansible_ssh_host=192.168.1.400 ansible_ssh_user=…
test2.lb.my.com ansible_ssh_host=192.168.1.401 ansible_ssh_user=…

[test1]
test1.db.my.com
test1.web.my.com
test1.lb.my.com

[test2]
test2.db.my.com
test2.web.my.com
test2.lb.my.com

So sometimes I need to run some update on all ‘webservers’, but sometimes I want to stop and reconfigure all servers in ‘test2’.

At last I need to address a specific environment and do things on different nodes iin sequence, e.g.

  • ping the ‘database’ inside
  • ‘df -h’ all the webservers
  • then ‘free -m’ all the load-balancers

How could I tell that the following playbook is only for ‘test1’ environment and I need to run the ‘databases’ part of the playbook only for the databases inside that environment?

What are the starting points or steps? Topics I should read?

Regards:
Bence

2

hosts: databases:&test1

^ will make the playbook target only hosts that exist in both groups
http://docs.ansible.com/intro_patterns.html#patterns

3

We followed the recommendation at

http://docs.ansible.com/playbooks_best_practices.html#stage-vs-production
and have separate inventory files for each distinct environment (in your case test1 and test2).

We then invoke ansible with “-i test1” or “-i test2” depending on what environment is being targeted. The ability to hit -all- webservers with one command is lost, but you usually don’t want to do that and it’s safer to update test1 before test2 (or production).

So far this has worked well.

4

You don't loose anything if you put both inventories in the same
directory, then you can call that directory as a 'global unified
inventory' with -i /dir/with/test1andtest2 and still address all
webservers.

5

Interesting!

So if, instead of having top-level files called “env1”, “env2” and “env3”, we created a “inv” folder and moved them all into there…
You’re saying we could run on just env1 with “-i inv/env1” or all envs with “-i inv”?

Will it still pull in the values from group_vars/env1 when we do the former?

6

Thanks, Brian

…and what if i just don’t want to declare it in the playbook?
Can I make the environment a parameter for the playbook?

hosts: “databases:&{{env}}”

or when calling the playbook:

ansible-playbook example.yml --limit test1

Regards:
Bence

  1. december 23., kedd 13:35:07 UTC+1 időpontban Brian Coca a következőt írta:
7

I found that:
http://stackoverflow.com/questions/18195142/safely-limiting-ansible-playbooks-to-a-single-machine

  1. december 25., csütörtök 16:41:17 UTC+1 időpontban Bence Takács a következőt írta:
8

Here is a detailed description of my original idea: http://rosstuck.com/multistage-environments-with-ansible/

It seems that the concept was not obvious for others either. But afer a little digging they come up with the same concept… :wink:

Regards:

Bence