Fork me on GitHub

Anyone prepared to share what they are doing about repeatability and configuration on AWS..? Clearly OpsWorks is already there, and that’s Chef by another name as far as I can tell, but it’s clear that Puppet and Ansible are both becoming or are already very AWS friendly… I am planning a personal development that I want to be able to orchestrate, but I am new to doing the DevOps Toolchain for myself and I am hoping someone is prepared to state a preference based on having walked this road ahead of me..?


@maleghast: we went with CloudFormation. more upfront work, but better long term, as it’s the most configurable. infrastructure as data.


we update and push github repo and circleci’s post-test hook invokes cfn and updates infra


our instances are very simple, so we didn’t need a whole abstraction for that. java + jar + monitoring and log-tail agents


@lowl4tency can answer in more detail simple_smile

Kira Sotnikov11:10:36

Yes, Robert are right in main line of our processes.

Kira Sotnikov11:10:27

maleghast: i suggest to use same tools as you use for main development. I suggest also to take a look on containers


as lowl4tency says try the containers thing (it was announced last week ECS), it uses CloudFormation beneath it

Kira Sotnikov13:10:28

I suggest try it locally first, if it's your first expirience

Kira Sotnikov13:10:39

and other tools by hashicorp such as consul, vagrant etc


I've been successful at making a clojure uberjar to run as a lambda function on AWS. Now trying to get it to post messages to SQS with not much success, but I'm still working on it simple_smile


@clojuregeek: does aws lambda preload the jvm for you? or do you have to count the jvm load time into your lambdas?


@sarcilav: i just choose java8 as my env and it works


they're doing something special, last i tried clojure + lambda it was quite snappy


yeah it seems to be good


mmm, nice to know, going to try later today


I have also made a lambda function with


@alandipert: Do you guys have a boot task for Clojure on Lambda? What about Castra and or Hoplon on Lambda


@tbrooke: we don't have a boot task for it, but castra on lambda is very interesting indeed


@alandipert: notice the lambada docs: "Simplest way to deploy is to create an uberjar using leiningen or boot " 👍


We’re using CloudFormation too. At first we thought it was the bees’ knees, but we’ve run into some problems as we’ve tried to refactor it. First, we ran into the 200-resource limit per stack. Now we’re refactoring into multiple stacks, but it’s somewhat tricky, because many of the resources need to be unique (think DNS settings etc), and CloudFormation wants to create new instances of those resources for the new stacks before it deletes them from the old stack, etc.