Ansible at large scale.

[iamgroot]

Wondering if there are any tips or leasons learned from fellow members who have used Ansible to manage/deploy bare metal compute systems numbering in the thousands.  For example, large HPC systems.

[Lisenet]

That's something I'd be interested in too. We currently use Foreman for provisioning, but I hear that people use Ansible for provisioning so wondering on how that works at an enterprise level.

[Rinaldi]

I'm also curious how it's implemented at an enterprise level with hundreds/thousands of servers (bare metals or virtual machines) with existing spacewalk for PXE boot, kickstart/updates etc.

What kind of playbooks do you use ? Is it one for base playbook (register to spacewalk, FreeIPA/LDAP, setup static IP, etc) and other playbooks based on the server's role i.e. database, web server, etc?

[ricardodacosta]

Are you guys interested in Infrastructure Automation? Just putting feelers out there ...

https://www.ansible.com/integrations/infrastructure

---------------------------------------------------------------------------------------------------------


If you're satisfied with the solutions provided please mark the solution as ACCEPTED.

Don't forget to thank those who helped you out with kudos!

[Lisenet]

@ricardodacostaWe are interested in people's experience with regards to Ansible automation at an enterprise level.

[ricardodacosta]

That's a very broad topic. And we always have a good number of customers talking about their experiences with Ansible at the Red Hat Summit every year, as well as at AnsibleFest.

Have a look at these links:

https://www.youtube.com/user/redhatsummit

https://www.youtube.com/channel/UCeImQ-jeVhzgLJd9wsqo8Sg/videos

If there is something specific that you'd like to hear more of, please message me.

[SeraKC]

I had experience using ec2.py for dynamic inventories with a rather... interesting setup for Ansible that was splitting out everything against a large amount of repos.

While it's not the recommended solution, I do think it offers some good insight in different ways to use Ansible.

Depending on the environment there could be a large number of machines defined in the configuration. It really depends on what platform you're using though, because if you're doing it on AWS you can deploy thousands of machines rather easily for the most part.

But if you're deploying thousands of machines via docker then the solution would be different.

From what I saw from a client here in the UK there are definitely BAD ways to do it where you can turn Ansible into a monolithic landscape that can become messy rather quickly. The key lesson I learned was to keep it as simple and condensed as possible, whilst still abstracting vars etc out.

If you're interested do feel free to shoot me a mail.