playing with ansible more
April 11, 2012
I’ve been working with/on ansible off and on for a couple of weeks now. I’ve got a simple reinstall playbook created and the basics of how we create new builders.
I was also working on using ansible as an api to port over some tools from func. After getting a bit sidetracked getting it to have a optparser function so I didn’t have to duplicate all those options in ever script I ported the func-host-reboot script over and improved it a bit:
https://github.com/ansible/ansible-contrib/tree/master/scripts/host-reboot
The advantage of using the python interface is I don’t have to write things using the playbook language when I really want to do something more involved and I like the idea of being able to do multiple commands BETWEEN hosts and interplay their results.
I’m thinking that, coupled with kickstart, we should be able to streamline all of our provisioning. Additionally, I’m thinking that the api (esp for the playbooks) should let me play with more ad-hoc builder creation and package build submission.
I’m trying to figure out where this fits.
We have pieces that can do all of these things, mostly, but all of them require some other kind of setup to make work. I have 3 f16 hosts, 1 el6 box and 1 el5 box that I’m testing these against. I run against all of them at the same time and, other than sshd and python none of them have any specific installed on them for ansible to be able to run.
Is a lightweight “clientless” mgmt system a good idea? Is it enough of a feature to make it worth pursuing?
It feels like it helps overcome the pain-in-the-ass quality that is setting up most systems-mgmt infrastructure.
January 15, 2013 at 9:25 am
So far, the lightweight “clientless” mgmt system looks to me like a good idea, compared to the likes of Puppet and Chef. It probably depends on your use case: I have small networks of up to several linux machines to configure and that’s about it. Ansible seems like the best tool for the job by far. There are, of course, scenarios where it doesn’t make sense (having to configure Windows machines, embedded devices etc.). What were your conclusions, since you published this 8 months ago?
January 15, 2013 at 8:17 pm
We’ve been using in fedora for a number of tasks and I think it is working quite nicely. I’ve never cared about windows machines as I don’t have anything like that in the infrastructure I work on. I also think that attempting to shoe horn windows into a tool which is originally developed to maintain unix-based systems is a recipe for creating a terrible tool for everyone.
I would discourage people from pushing ansible to support windows – that’s just a recipe for an ugly program ill-fitted to anything.
January 15, 2013 at 9:14 pm
I’m glad it’s working out for you. I’m not interested in Windows either, but thought I’d mention to illustrate the boundaries of ansible’s scope to anyone else who might be reading.