/test-echo-service

Simple echo service

Primary LanguageHCL

test-echo-service

Requirements

  • working AWS account - http://aws.amazon.com/

  • AWS user with API access and rights to create keys and EC2 instance - can be created inside AWS IAM, we will need access key and secret key

  • aws cli tool installed

    • brew install awscli on MacOS X,
    • apt-get install awscli on Ubuntu/Debian,
    • pip install awscli on CentOS,

    please see https://docs.aws.amazon.com/cli/latest/userguide/installing.html for more information

  • Docker daemon installed, Dockerhub username and password - https://www.docker.com/

  • terraform installed - https://www.terraform.io/

  • make utility - optional, you can just execute commands from Makefile, brew install in MacOS X, apt-get install make in Debian/Ubuntu, yum install make in CentOS

  • jq utility - optional, pretty prints JSON output

Steps to reproduce / make it work

  • Clone git repo or download it - git clone https://github.com/mvasilenko/test-echo-service

  • cd src

    • docker login - enter your Dockerhub username and password
    • docker build -t dockerhub_username/test-echo-service . - this will build Docker image locally in current . directory, don't miss . at the end of the line
    • docker push dockerhub_username/test-echo-service - this will push (store) newly built image to Dockerhub, so it can be pulled from remote location (EC2 instance in our case)

  • cd deploy

    • configure custom aws cli profile

      • aws configure --profile test-echo-service - enter your AWS access key and secret key OR
      • edit $HOME/.aws/config and $HOME/.aws/credentials

      This step is needed because Amazon Route53 metrics available only in us-east-1 region, alert email subscription initiated by executing aws command localy, as stated in deploy/sns-r53-healthcheck.tf, email subscription must be confirmed, by clicking on link in email from AWS CloudWatch

      Please see hashicorp/terraform-provider-aws#712 for more information.

    • edit terraform.tfvars file, place AWS and Docker credentials here, this is private file, DON'T share it

    • review vars.tf - default variables, edit your email at last lines

    • terraform init - download and init terraform providers

    • terraform plan - evaluate provisioning manifests, dry-run

    • terraform apply -auto-approve - make changes, provision infrastructure, it will output EC2 instances public IP

    • go to AWS Console, confirm email alert subscribe

    • check the service, you will see such output if everything is OK

$ echo -n '[17/06/2016 12:30] Time to leave' | nc 52.59.245.195 1234 | jq .
{
  "timestamp": 1466166600,
  "hostname": "host.local",
  "container": "host-container-1",
  "message": "Time to leave"
}
  • play with it - ssh -l ubuntu ec2_instance_ip from previous step
    • docker ps - to see running containers
    • docker logs host-container-1 -f - container logs
    • docker stop host-container-1 - stop echo service
    • check email for alert from AWS CloudWatch
  • terraform destroy -force - destroy infrastructure when you done, to prevent extra AWS charges

Files description

src/Dockerfile - rules for docker image build src/echo-server-asyncio.py - echo server

Container with service started with cloud-init at instance boot. Runtime parameters to docker container (HOSTNAME, CONTAINERNAME, PORT) are passed as environment variables.

deploy/*.tf - terraform provisioning manifests deploy/files/cloud-init.cfg - Cloud-init template

What can be improved or done in other way

  • make docker image private - now it's public, so passing dockerhub username/password to cloud-init is pretty useless
  • send email in event of a service recovery (not done)
  • store terraform state at S3
  • store terraform AWS credentials in $HOME/.aws instead of terraform.tfvars
  • better handling for cloud-init steps - now all steps (install docker daemon, docker login, docker run) are repeated every time at every instance boot
  • remote docker build instead of local - we can pack docker build command into cloud-init configs, this will get rid of Dockerhub and push/pull operations
  • versioning for image - embed git describe or equivalent in docker image tag
  • implement load balancing and AWS auto scaling, triggered by health check