FizzBuzz in Kubernetes

A distributed, fault-tolerant, production-ready FizzBuzz implementation for the microservice cloud.

Run the FizzBuzz

Dependencies:

• aws cli (pip install --upgrade --user awscli)
• jq (brew install jq)
• kops (brew install kops)
• kubectl (brew install kubectl)
• An existing Route53 Hosted Zone that is correctly serving DNS queries.
  • Easiest way to do this is simply set up a Hosted Zone for a subdomain already handled in Route53. Something like cluster.example.com.
• aws cli must be authenticated with a user that has the following group policies
  • arn:aws:iam::aws:policy/AmazonEC2FullAccess
  • arn:aws:iam::aws:policy/AmazonRoute53FullAccess
  • arn:aws:iam::aws:policy/AmazonS3FullAccess
  • arn:aws:iam::aws:policy/IAMFullAccess
  • arn:aws:iam::aws:policy/AmazonVPCFullAccess

Then, finally,

./bin/get-programming-job.sh cluster.example.com

What Happens

• Create an S3 bucket to store the kops state,
• Create a multi-zone Kubernetes cluster using kops,
• Wait for the cluster to spin up and schedule pods,
• Deploy an ElasticSearch StatefulSet,
• Deploy a fluentd DaemonSet running a pod on each node in the cluster which forwards logs to ElasticSearch for indexing,
• Deploy a 100-replica StatefulSet of the fizzbuzzer container, which accepts a StatefulSet hostname (fizzbuzzer-n) as an argument and computes whether or not it should output Fizz, Buzz, FizzBuzz, or n.
• Waits for all of those pods to have spun up and then,
• Queries the ElasticSearch Search API to collect all of the logs from the fizzbuzzer pods,
• Outputs the result of the FizzBuzz,
• Tears everything down (cluster, s3 bucket)

Why

Someone had to.

Issues

Occasionally there are DNS and connectivity issues, but I'm just going to ignore them because I've spent enough time on this as it is and it works most of the time.