Serverless Task Scheduler for AWS

Using AWS DDB, SQS and Lambda.

Task Types

Supported task types functions should be exported by ./lambda/taskRunners/index.js.

Task runners receive a ddbTask as an argument and should return an object like this:

{
  "completed": true,
  "newParams": {},
  "newTargetTimestamp": null
}

• completed: wether the task has been completed in this iteration or not.
• newParams: params to update in the DDB item.
• newTargetTimestamp: timestamp (in ms) at witch the next iteration of the execution should be executed.

See ./lambda/taskRunners/ for an example.

Setup

# Clone the repository
git clone git@github.com:imharvol/serverless-task-scheduler-aws.git
cd ./serverless-task-scheduler-aws/

cd ./lambda/

# Add task types to ./taskRunners/

# Bundle the function into a .zip
npm run lambda-symlink && npm run bundle

# Bundle the dependencies into a .zip (this will be the layer)
# Feel free to use my package create-lambda-layer-aws, but you can also zip them manually
npx create-lambda-layer-aws ./package.json

cd ../

# Move the function bundle and layer bundle into the terraform folder
mv ./lambda/function.zip ./terraform/
mv ./lambda/layer.zip ./terraform/

# Initialize the terraform project
terraform init

# Apply the terraform configuration
terraform apply

terraform apply will output a tasks_queue_url and a tasks_table_name which you will need to send tasks. See example.js for an example of how to create and execute a task.

DDB Task Items

DDB Task Items should have the following format:

{
  "id": 0,
  "completed": false,
  "type": "exampleTask",
  "params": {}
}

• id: task id, it can be generated using node:crypto's randomUUID function.
• completed: wether the task is completed or not, this should be initialized as false.
• type: task type name, this task type should be exported by ./lambda/taskRunners/index.js.
• params: any other parameters needed in the task.

SQS Task Messages

SQS Task Messages should have the following format:

{
  "taskId": 0,
  "targetTimestamp": null,
  "priority": null
}

• taskId: task id in DDB.
• targetTimestamp: timestamp (in ms) at which we want the task to be executed.
• priority: reserved for future use.