Skip to content

Latest commit

 

History

History
213 lines (147 loc) · 7.29 KB

README.md

File metadata and controls

213 lines (147 loc) · 7.29 KB
Raw

Ansible Playbook for Serverless Framework

Example Ansible playbook for deploying Serverless service.

Benefits on using this approach:

  • The service can be tested in the build flow
  • Stack deployment is faster when the service is already compiled and test are runned
  • Same tested package is used in all the stages

For small scale serverless projects there is also a lite version that builds and deploys Serverless service serverless-deployment-ansible-lite.

Contents

Project Structure

Setup Local Environment

Setup Jenkins

Deployment Flow

Build

Deployment

Project Structure

  • group_vars default variables
  • inventories inventory files and variables for environments (development, production, etc.)
  • roles
    • infra role for infrastructure, vpc, database etc.
    • service role for Serverless service
  • scripts scripts that helps deployment

Setup Local Environment

If you are deploying from your local environment, Docker is required for deploying the playbook or alternatively Ansible and all the dependencies defined in Dockerfile installed in you environment.

Setup Jenkins

When using jenkins for deployment, easiest way is to setup Jenkins into EC2 instance running in your AWS account. Here is quick and dirty instructions how to do that Jenkins setup instructions

In addition to suggested plugins, install following plugins also:

  • Pipeline: AWS Steps
  • Version Number Plug-In

Deployment flow

deployment flow

  1. Create artifact from the Serverless service that contains parameterized CloudFormation Jinja2 template and functions package
  2. Create artifact from other services, e.g. website frontend
  3. Upload artifacts to S3 (or Artifactory or other file system).
  4. Download artifact with Ansible
  5. Extract and create environment specific CloudFormation templates
  6. Deploy stack to AWS

Build

Create S3 bucket for artifacts using AWS Console or aws-cli.

aws s3api create-bucket --bucket my-artifacts-bucket --region us-east-1

Local environment

serverless-deployment-example-service has ready made setup for serverless-ansible-build-plugin that helps the artifact creation process.

In the serverless-deployment-example-service, scripts directory there is create-artifact.sh that builds service and creates artifact with recommended filename.

Executing create-artifact.sh with param -p creates tar package, -s parameter is mandatory and needs to match service name.

Following snippet will create package named e.g. example-service-1.20170206.1-799dcd4.tar.gz. Where

  • example-service is service name (set with -s)
  • 1 is version for service (set with -v, optional, defaults to 1)
  • 20170206 is build date
  • 1 is number of the build in build date (set with -b, optional, defaults to 1)
  • 799dcd4 is short hash from git
./scripts/create-artifact.sh -p -s example-service

The 1.20170206.1-799dcd4 part is the service_version what is used later in Ansible vars.

Then copy the artifact to you artifact S3 bucket.

Jenkins

Following pipeline script will

  1. checkouts the service from repository
  2. creates version tag
  3. builds service
  4. uploads artifact to S3 bucket
MAJOR_VERSION = "1"
ARTIFACTS_BUCKET = "my-artifacts-bucket"
SERVICE = "example-service"
node {
    stage('Checkout service repo') {
        dir('project') {
            git 'https://github.com/SC5/serverless-deployment-example-service.git'
        }
    }
    
    stage('Set version') {
        dir('project') {
            gitCommitShort = sh(returnStdout: true, script: 'git rev-parse --short HEAD').trim()
            version = VersionNumber(projectStartDate: '2016-10-01', skipFailedBuilds: true, versionNumberString: '${BUILD_DATE_FORMATTED, "yyyyMMdd"}.${BUILDS_TODAY, X}-${gitCommitShort}', versionPrefix: "${MAJOR_VERSION}.")
            currentVersion = "${version}" + "${gitCommitShort}"
            
        }
    }
    
    stage('Build') {
        dir('project') {
            withEnv(["VERSION=${currentVersion}"]) {
                sh "./scripts/create-artifact.sh -s ${SERVICE}"
            }
        }
    }
    
    stage("Copy Artifact to S3") {
        dir('artifacts') {
            withEnv(["VERSION=${currentVersion}"]) {
                sh "truncate latest --size 0 && echo ${currentVersion} > latest"
                sh "tar -zcf ${SERVICE}-${VERSION}.tar.gz -C ../project/.ansible ${SERVICE}.zip ${SERVICE}.json.j2"
                withAWS {
                    s3Upload(file:"${SERVICE}-${VERSION}.tar.gz", bucket:"${ARTIFACTS_BUCKET}", path:"${SERVICE}/${SERVICE}-${VERSION}.tar.gz")
                    s3Upload(file:"latest", bucket:"${ARTIFACTS_BUCKET}", path:"${SERVICE}/latest")
                }
            }
        }
    }
}

Deployment

Define artifacts S3 bucket to artifacts_bucket variable in group_vars/aws/vars.yml, so that Ansible knows where to download artifacts.

Local environment

First define the service name to roles/service/tasks/main.yml

- set_fact:
    service_name: example-service

When deploying from local environment AWS secrets needs to be passed to deployment container, for that e.g. .deploy.sh script in the project root with contents of

#!/usr/bin/env bash

export AWS_ACCESS_KEY=my-access-key
export AWS_SECRET_KEY=my-secret-key

./scripts/deploy-local.sh

might ease up the deployment flow.

  1. Build Dockerfile with ./scripts/build-docker.sh
  2. Run ./.deploy.sh

To deploy certain version of the Serverless service, use environmental variable SERVICE_VERSION. If version is not defined it fallbacks to latest version.

Jenkins

When using Jenkins on AWS EC2, the role of the instance needs to have permissions to deploy CloudFormation stacks, create S3 buckets, IAM Roles, Lambda and other services that are used in Serverless service.

Following pipeline script will

  1. checkout this repository
  2. build docker image for deployment
  3. deploy stacks to AWS
node {
    stage('Checkout repository') {
        git 'https://github.com/SC5/serverless-deployment-ansible.git'
    }
    stage('Build Docker image') {
       sh "./scripts/build-docker.sh"
    }
    stage('Deploy') {
        sh './scripts/deploy-development.sh'
    }
}

The deploy script ./scripts/deploy-development.sh in deployment repository executes Ansible in docker container.

#!/usr/bin/env bash

docker run -v $(pwd):/src -w /src serverless-deployment /bin/bash -c "\
export SERVICE_VERSION=$SERVICE_VERSION && \
ansible-playbook -vvvv --inventory-file inventories/development/inventory site.yml"

[[ $? -eq 0 ]] || { echo "Deployment failed!" ; exit 1; }

To deploy certain version of the Serverless service, use environmental variable SERVICE_VERSION, that can be defined in Jenkins parameterized job. If version is not defined it fallbacks to latest version.