We’re living in interesting times. Till 4–5 years ago, infrastructure used to be locked by review boards and tickets. Open source has been slowly changing this world. Take OpenStack, Docker, Mesos, Kubernetes, you name it.

But here is a message for everyone jumping into these. If you don’t understand manageability you will fail. Failure won’t happen immediately, but you will go through some phases over a period of time.

I’ll give a few simple reasons:

• Infrastructure is brittle. It takes a lot of automation to make it malleable.
• Most infrastructure software is distributed. Understanding their failure modes and designing for failure are non-trivial exercises.
• It takes a lot of software to manage these systems at scale, and you need to apply software engineering to operational tasks.
• There is no free lunch.

Don’t pick X, where X is any of the above tech, just because some name brand company picked it or event built it. Pick X only if you’re prepared for the long haul and ready to build a ton of software to manage X.

Originally published at https://www.subbu.org/blog/2015/11/its-the-manageability-stupid.