On Immutable Infrastructure

I believe there’s little convincing to be done on the merits of immutable deployment. With immutable deployment, the unit of abstraction is a machine image; a docker image, Amazon Machine Image (AMI), e.t.c. The idea is that your server, spawned from a machine image goes from being this special snowflake to [insert generic thing]. As opposed to meticulously maintaining a server and keeping track of its mutable configuration over time, you are able to provision it with a known configuration in a non-special way. Containerisation makes this easier because of the lightweight and wieldy characteristics of containers.

Last week, I listened to a talk on immutable infrastructure, titled Immutable Infrastructure: Rise of the Machine Images. The speaker expands on immutable deployment by talking about how certain ancillary aspects of applications are affected by immutability, such as logging, sessions, configuration and service discovery. I highly recommend that talk as the speaker puts into words vaguely defined practices that have emerged from this growing trend.