top | item 37332528

(no title)

planetafro | 2 years ago

I do understand your point but would counter that, if done right, you can revert your last commit. :)

discuss

order

anyoneamous|2 years ago

Sadly it's not that simple. For example:

- Run a playbook which asserts the existence of a user

- Playbook fails at a later stage

- Executing an earlier version of the playbook which doesn't have the user assertion won't cause the user to be removed, just ignored. You need a different (assert not-existing) statement to actually remove the user.

In general these things won't be impactful - but they do mean that if you iterate on your deployment playbooks, you will end up with orphaned files/resources scattered around unless you explicitly tidy them up.

IME using Ansible for small-scale automation is no faster than just running the tasks manually, once development time is taken into account - you need tens of servers to make it worthwhile, by which point containers start to look more appealing.

okeuro49|2 years ago

> IME using Ansible for small-scale automation is no faster than just running the tasks manually, once development time is taken into account - you need tens of servers to make it worthwhile, by which point containers start to look more appealing.

Ansible has a very low learning curve. I cringe with the thought of the time it would take documenting then typing in all the commands to setup my servers manually. That's just not feasible. Ansible provides a sweet spot between traditional "pet" servers and containerisation.

It's not only the number of servers that matters. It's also about being able to create ephemeral staging/developer environments, and easily being able to easily migrate to different VM providers.

listenallyall|2 years ago

> IME using Ansible for small-scale automation is no faster than just running the tasks manually, once development time is taken into account

You aren't wrong here but still the payoff is typically worth it, especially for small scale. "Development time" is one factor but the alternative -- building servers or infrastructure manually -- causes all kinds of problems when individual servers are not consistent.