AWSBoxen Deployment Config for picl-idp
=======================================
This directory specifies how to deploy picl-idp into AWS. It's a templated
CloudFormation stack with some extra details about how to build AMIs. You
will need "awsboxen" to interpret and act on these instructions:
https://github.com/mozilla/awsboxen
Most of the interesting stuff is in the "Resources.yml" file, where we
declare all the various resources needed in the deployment.
At a high level, the deployment consists of the following:
* An auto-scaled group of *webheads* running the picl-idp nodejs
application. These are stateless machines which defer all their
data-storage needs to...
* An auto-scaled group of *cassandra* nodes. These act as a dumb
data store and have a bit of trickery to auto-configure themselves
into a cluster.
* Everyone sends logs to a stand-alone *log aggregator* using heka.
This runs an elasticsearch+kibana interface for analysing the logs,
view graphs, etc.
Things to do:
- webheads:
- local postfix relaying to SES
- cassandra:
- auto-clustering magic
- add more vnodes, configure auto-bootstrapping, etc
- VPCs and subnets
- Use existing bastion hosts, lock down ssh to bastion hosts only
898bb68df0