Skip to content

kagarlickij/aws-cloudformation-ec2-r53

Repository files navigation

Build status License
CircleCI License: MIT

Task

Make FreeRADIUS installation highly available

Solution

RADIUS uses UPD ports so LoadBalancer doesn't suite

Each instance has cron job that checks radiusd service status each minute and sends data to CloudWatch metric

If radiusd service on first instance is not active or no data was received CloudWatch alarm status will be changed from OK to ALARM

Route53 health check is listening to CloudWatch alarm and will become unhealthy as soon as CloudWatch alarm is

Traffic will be redirected to second instance by Route53 Failover Routing Policy until first instance is back

Both instances will be automatically replaced in case of termination by ASG and LaunchTemplate that uses cfn-init scripts to install FreeRADIUS

diagram

Known issues and workarounds

  1. Route53 Record set values are added in random order, this means that by default traffic could be sent to second instance

    Temporary solution is to stop radiusd service on first instance, wait for alarm and than start service

    This will make first instance default target for traffic

    Permanent fix to implement: Route53 Record Set should be removed from CloudFormation and implemented as CI step using either CLI or SDK

  2. Route53 Traffic policy is more advanced than routing policy in record, but it's not available in CloudFormation

  3. Solution doesn't configure FreeRADIUS, only installs it

Production usage caution

This code is for demo purposes only and should never be used in production.

About

CloudFormation template with CI to deploy EC2 instances with Route53 health check

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages