What is most valuable?
The beauty of Ansible is the easy ramp-up to get started. You really only need Python and SSH access. Configuration is generally done in YAML, which is easy to understand, and there is a progression from ad hoc tasks, to playbooks, then to roles, which means you can start with one server and continue building up to datacenters worth of servers with the same methodology. Also, shared by most configuration management tools, the idea of creating a desired state scales better than trying to specify procedural steps to set up new hosts. There are no agents by default, so adding a new server is a matter of a couple lines of configuration (on a new server and the configuration master).
How has it helped my organization?
There is some overhead in setting up the initial playbooks, but it now takes less time to set up 10 servers than it did to configure one in the past. Also, the setup is consistent because there is not the concern that someone forgot to copy/paste a config line or run another command. Whatever is in the playbook gets done.
What needs improvement?
Because Ansible is establishing SSH sessions to perform tasks, there is a limit on scalability. Speed and the sheer number of open connections start to become issues past a couple hundred servers. There are some workarounds, but that is a key area for improvement. Ansible could also improve support for private package repos, to ensure that new batches of servers are getting the same package versions as earlier batches.
For how long have I used the solution?
I have been using Ansible for about two years.
What do I think about the scalability of the solution?
SSH is pretty good, but it was not designed for the access pattern of hundreds of connections out of configuration targets. Other tools solve this with a listening agent process, so the initial connection to configure is much faster.
How are customer service and technical support?
I have not used customer service. Ansible is well established, so there is plenty of documentation, examples, and third-party resources.
Which solution did I use previously and why did I switch?
Manual configuration and "Golden" templates for virtual machines were used. The former is tricky to maintain consistency with. The latter seemed to require constant updating and it did not help maintain the configuration of already installed servers.
How was the initial setup?
Initial setup boils down to installing Ansible and ensuring you have SSH access to a target that is running Python. Standard packaging is available on major Linux distros to install some level of Ansible. I recommend following instructions on Ansible's site to get the latest stable release as they have been improving rapidly.
What was our ROI?
What's my experience with pricing, setup cost, and licensing?
Although Red Hat has an enterprise add-on to manage Ansible through a web application and offers commercial support, I have not used it. Like many Red Hat products, they have a no-cost version of the web application (AWX, formerly Ansible Tower), but you are on your own to install and it is a little more complicated than just installing Ansible. AWX will probably be required in most shops for the RBAC functionality. With AWX, non-admins can be limited to perform some tasks, but not be allowed free reign with Ansible.
Which other solutions did I evaluate?
Salt (or SaltStack) is a similar tool, but does have an agent. There are other tools like Chef or Puppet that use languages other than Python. Ansible was chosen based on these characteristics and the others were not evaluated after this initial choice.