AWS Lambda function to stop and start EC2 instances based on resource tag using crontab-like expressions
Usage
Testing and development
Creating a Lambda Deployment Package
Build environment
Serverless build pipeline
Identity and Access Management policy
Creating and scheduling Lambda function
Lambda function looks for EC2 instances and Auto Scaling Groups that has a resource tag ec2Powewrcycle attached to it.
Tag value is simple JSON document that describes start and stop schedule in crontab-like expressions.
In case of ASGs, the tag may also contain information about the scaling state of the group (min and desired instances in the group).
If it doesn't, then the min and the desired instances are defaulted both to 1.
-
EC2 instance stop/start schedule: Mon - Fri, 8.00am - 5.55pm
asgLifecycle: { "start": "0 8 * * 1-5", "stop": "55 17 * * 1-5" }
-
Auto Scaling Group stop/start schedule: Mon-Fri, 9:00am - 11:00pm. Min no of instances in ASG is 2 and desired no is 3
asgLifecycle: { "start": "0 9 * * 1-5", "stop": "00 23 * * 1-5", "min": 2, "desired": 3 }
-
Auto Scaling Group without scaling state specified. This will default to Min no of instance in ASG to 1 and the desired no to 1
asgLifecycle: { "start": "0 8 * * 1-5", "stop": "55 17 * * 1-5" }
-
Auto Scaling Group without min state specified. This will default to Min no of instance in ASG to 1.
asgLifecycle: { "start": "0 8 * * 1-5", "stop": "55 17 * * 1-5", "desired": 3 }
As of commit 00389de the stop/start schedule can be defined as an URL to publicly accessible JSON document. This feature can be handy when managing schedule for large number of nodes.
ec2Powercycle: https://raw.githubusercontent.com/Financial-Times/ec2-powercycle/master/json/dev-schedule.json
To run ec2Powercycle job local dev environment you need to install all dependencies such as boto3 and croniter. Full list of dependencies can be found in the file ec2_powercycle.py
You also need to set up AWS credetials (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY and AWS_DEFAULT_REGION) in order to interact with AWS API.
To run the job first change to python directory inside the repository, then call the hander() function.
cd /path/to/repository
python -c "from ec2_powercycle import * ; handler()"
Function can be executed in so-called dryrun mode with the following command.
python -c "from ec2_powercycle import * ; handler({ \"DryRun\": \"True\" })"
In dryrun mode function doesn't stop/start instances.
EC2-POWERCYCLE uses 3rd party library called Croniter which must be installed before deployment package is created.
pip3 install croniter requests -t lib/
The following command is run in the root of the ec2-powercycle repository. The command bundles ec2-powercycle business logic, its dependencies and the README.md which can be uploaded to Lambda or S3 bucket.
zip -r ../ec2-powercycle-0.0.1.zip ./*.py lib/ README.md
This repository ships with Dockerfile that can be used for packaging and deployment automation.
The following command is run in the root of the repository and it creates a Docker container called ec2-powercycle with tag value 1.
sudo docker build -t ec2powercycle .
When Docker image is running it first executes the packaging script package.sh, then deployment script lambda-deploy-latest.sh that pushes ec2-powercycle.zip package into Lambda.
To run lambda-deploy-latest.sh in headless mode you can provide AWS credentials as Docker environment variables.
sudo docker run --env "AWS_ACCESS_KEY_ID=<access_key_id>" \
--env "AWS_SECRET_ACCESS_KEY=<access_key_secret>" \
--env "AWS_DEFAULT_REGION=<aws_region_for_s3_bucket>" \
--env "AWS_LAMBDA_FUNCTION=<lambda_function_name>" \
-it ec2powercycle
Launching Docker image without environment variable will run post-to-lambda.sh in interactive mode that prompts user for AWS credentials.
sudo docker run -it ec2-powercycle
Circleci is a hosted CI service that integrates nicely with Github and AWS.
Build pipeline currently has a single workflow with three possible tasks: deploy-test
, deploy-master
and deploy-release
.
sequenceDiagram
autonumber
Note right of developer: deploy-test
developer-->>github: git push feature/branch
Note right of CircleCI: Dev account
CircleCI-->>github: git clone feature/branch
CircleCI-->>AWS: Upload the function
CircleCI-->>AWS: Invoke the function in "dryRun" mode
Note right of developer: deploy-master
developer-->>github: merge pull request
CircleCI-->>AWS: Create version for the last uploaded function
CircleCI-->>AWS: Create alias(named LIVE) for that version
Note right of developer: deploy-release
developer-->>github: git push tag release-*
Note right of CircleCI: Prod account
CircleCI-->>github: git clone release-*
CircleCI-->>AWS: Upload the function
CircleCI-->>AWS: Create version for the last uploaded function
CircleCI-->>AWS: Create alias(named LIVE) for that version
The Development task is run every time the master branch is updated. Development task creates a deployment package, deploys it to Lambda and invokes the function against DEV alias.
Once you have completed Development work and wish to "promote" your code to Production you can trigger Production task by creating a Git tag with prefix release- and pushing the tag to repository.
Use the following commands to create a tag and push it to repository.
git tag -a release-12 -m "Repoint LIVE alias to release-12 tag"
git push origin release-12
AWS credentials are already added to two different contexts ec2-powercycle-test and ec2-powercycle-prod as environment variables. The workflow invokes the corresponding context with its AWS credentials in order to deploy deployment package to Lambda.
To change credentials:
- Go to Circleci Contexts and click on ec2-powercycle-test or ec2-powercycle-prod
- Delete the old environment variable
- Create new environment variable
When creating Lambda function you will be asked to associate IAM role with the function.
The following policy example enables Lambda function to access the following AWS services:
- CloudWatch - Full access to Amazon CloudWatch for logging and job scheduling
- EC2 - Access to query status and stop/start instances when resource tag ec2Powercycle is attached to the instance and environment tag does not equal p (p=production)
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": [
"logs:CreateLogGroup",
"logs:CreateLogStream",
"logs:PutLogEvents",
"logs:DescribeLogStreams"
],
"Resource": "arn:aws:logs:*:*:*"
},
{
"Effect": "Allow",
"Action": [
"ec2:Describe*",
"autoscaling:Describe*"
],
"Resource": "*"
},
{
"Effect": "Allow",
"Action": [
"ec2:StartInstances",
"ec2:StopInstances"
],
"Condition": {
"StringLike": {
"ec2:ResourceTag/ec2Powercycle": "*"
},
"StringNotEqualsIgnoreCase": {
"ec2:ResourceTag/environment": "p"
}
},
"Resource": "arn:aws:ec2:*:*:instance/*"
},
{
"Effect": "Allow",
"Action": [
"autoscaling:UpdateAutoScalingGroup"
],
"Condition": {
"StringLike": {
"autoscaling:ResourceTag/ec2Powercycle": "*"
},
"StringNotEqualsIgnoreCase": {
"autoscaling:ResourceTag/environment": "p"
}
},
"Resource": "*"
}
]
}
The following policy enables build and deployment job to update Lambda function, invoke it and update aliases.
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": [
"lambda:CreateAlias",
"lambda:GetFunction",
"lambda:InvokeFunction",
"lambda:List*",
"lambda:PublishVersion",
"lambda:UpdateAlias",
"lambda:UpdateFunctionCode"
],
"Resource": [
"arn:aws:lambda:*:*:function:ec2-powercycle"
]
}
]
}
Once deployment package has been created we can create a Lambda function and use CloudWatch to set the function to run periodically.
-
Log on to AWS console and go to Lambda configuration menu
-
Click Create a Lambda function
-
In Select blueprint menu choose one of the blueprints (e.g. s3-get-object-python) click Remove button on the next screen to remove triggers. Then click Next.
-
on Configure function page provide the following details:
- Name*: ec2-powercycle
- Description: Optional description of the function
- Runtime*: Python 3.7
-
In Lambda function code section select Upload a .ZIP file to upload ec2powercycle.zip package to Lambda
-
In Lambda function handler and role section set handler name ec2_powercycle.handler
-
Select the role that has the above IAM policy attached to it
-
Set Timeout value 1 min
-
Click Next and Create function
- In Lambda configuration menu open the ec2-powercycle Lambda job
- Go to Triggers tab
- Click Add trigger
- Select Event source type: CloudWatchEvents - Schedule and provide the following details
- Rule name: whatever unique name
- Rule description:optional description of the rule
- Schedule expression: rate(15 minutes)
- Click Submit to create schedule