A tool for generating a simple visualization of AWS EC2 usage allocations.
AWS's Reserved Instance Billing can be a little confusing for some; Reserved Instances are cheaper, but it also requires knowing exactly what type of instance type is needed and where it resides.
As EC2 resources are added, removed, and shuffled, it can become easy to lose track of what's where and think you're getting the benefits of Reserved Instances when in fact you're not. Or, quite the opposite, you're using too many Spot Instances and could be benefiting from Reserved Instances but are not.
This script (which is still in early development) generates a visual report of how instances are allocated, allowing quick visual inspection of how to best reallocate servers, especially across availability zones.
IMPORTANT NOTE: I use this script on an EC2 instance with an AMI Role, so there's no AWS secrets baked into the script.
SHORTCOMING: Amazon recently announced a new class of Reserved Instances called Regional, which will provide discounted prices regardless of availabilty zone. This implementation does not know how to handle the data (yet?); using it will manifest as an 'undefined' column showing the correct number of reserved instances in grey, but it will not show spot-instances in other availability zones as if they were in the Regional column. I'm pondering a good way to represent both location and costing, as they can now be in two different columns simultaneously.
Instructions for installing NodeJS by source or package are online. This version was tested with Node v6.9.4.
$ git clone git@github.com:wls/aws-reserved-instance-allocation-report.git
$ cd aws-reserved-instance-allocation-report
$ npm install # Gets all the modules (do this only once)
$ node ri-report.js > report.html # Generate a report (assumes us-east-1)
$ firefox report.html # View the report
$ node ri-report.js us-west-2 > report.html # Specify us-west-2 as the desired region
$ firefox report.html # View the report
$ ./reports.sh us-east-1 us-west-2 eu-west-1 ap-northeast-1 # Run a lot of regions
$ ./reports.sh # Run default of us-east-1 and us-west-2
Reports are generated in the form of YYYY-MM-DD-region_name.html.
The script s3-reports.sh may be used to send the reports to S3.
Right now, that location is hard coded in the script so you will
need to alter the script for your own location (it's a the top).
This is one of those cases where personal need hasn't required
greater flexibiliy.
Likewise, these report scripts have a subset of aws regions that you're encouraged to modify, as the use case of needing all regions is fairly rare.
The code was written in JavaScript, with the intent of development on NodeJS. However, the long term view is that it might run in the browser, or optionally as a Lambda service that detects when nodes are added, removed, started, or shutdown, and generates a report that might even be mailed to an admin.
This code was quickly whipped out to scratch an personal itch, but seemed to provide enough utility worth sharing as I'm sure other people are in the same boat, especially when multiple people are managing AWS resources.
The system makes use of the AWS SDK, and requests a listing of all the reserved instances that have been purchased. This identifies the reservations for all availablity nodes across all the various instance types.
Then the system does the same thing to inspect the running instances to get actual counts.
Between both lists, it's possible to compute the super-set of all declared availabilty zones, reserved or in use, as well as the same for the instance types.
The system then meshes to two data structures against what it has and what it could have; the resulting analysis is then presented visually.
Green boxes are efficiently utilizing reserved instances. Grey boxes are reserved instances that are not in use. And Red boxes are instances for which there could be a reserved instance if needed (to get lower pricing).
The ideal strategy is the pack as many of the red boxes (overuse) into the grey ones (underrun).