From 3654a4848d3d4b274c509cfaf376c112f63b7dbf Mon Sep 17 00:00:00 2001 From: Piradeep Kandasamy Date: Tue, 30 Apr 2019 14:42:49 -0700 Subject: [PATCH] Design proposal for Queue Worker Service --- design/aws-ecs-autoscaling-queue-worker.md | 225 +++++++++++++++++++++ 1 file changed, 225 insertions(+) create mode 100644 design/aws-ecs-autoscaling-queue-worker.md diff --git a/design/aws-ecs-autoscaling-queue-worker.md b/design/aws-ecs-autoscaling-queue-worker.md new file mode 100644 index 0000000000000..3df852dc3f5ae --- /dev/null +++ b/design/aws-ecs-autoscaling-queue-worker.md @@ -0,0 +1,225 @@ +# AWS ECS - L3 Construct for Autoscaling ECS/Fargate Service that Processes Items in a SQS Queue + +To address issue [#2396](https://github.com/awslabs/aws-cdk/issues/2396), the AWS ECS CDK construct library should provide a way for customers to create a queue worker service (an AWS ECS/Fargate service that processes items from an sqs queue). This would mean adding new ECS CDK constructs `Ec2QueueWorkerService` and `FargateQueryWorkerService`, that would take in the necessary properties required to create a task definition, an SQS queue as well as an ECS/Fargate service and enable autoscaling for the service based on cpu usage and the SQS queue's approximateNumberOfMessagesVisible metric. + +## General approach + +The new `ecs.QueueWorkerServiceBase`, `ecs.Ec2QueueWorkerService` and `ecs.FargateQueueWorkerService` classes will create L3 constructs for: + +* Ec2QueueWorkerService +* FargateQueueWorkerService + +A `QueueWorkerService` will create a task definition with the specified container (on both EC2 and Fargate). An AWS SQS `Queue` will be created and autoscaling of the ECS Service will be dependent on both CPU as well as the SQS queue's `ApproximateNumberOfMessagesVisible` metric. + +The `QueueWorkerService` constructs (for EC2 and Fargate) will use the following existing constructs: + +* Ec2TaskDefinition/FargateTaskDefinition - To create a Task Definition for the container to start +* SQSQueue - The queue that the worker is processing from +* Ec2Service/FargateService - The Service running the container + +## Code changes + +Given the above, we should make the following changes to support queue workers on ECS (for both EC2 and Fargate): +1. Create `QueueWorkerServiceBaseProps` interface and `QueueWorkerServiceBase` construct +2. Create `Ec2QueueWorkerServiceProps` interface and `Ec2QueueWorkerService` construct +3. Create `FargateQueueWorkerServiceProps` interface and `FargateQueueWorkerService` construct + +### Part 1: Create `QueueWorkerServiceBaseProps` interface and `QueueWorkerServiceBase` construct + +The `QueueWorkerServiceBaseProps` interface will contain common properties used to construct both the Ec2QueueWorkerService and the FargateQueueWorkerService: + +```ts +/** + * Properties to define an Query Worker service + */ +export interface QueueWorkerServiceBaseProps { + /** + * Cluster where service will be deployed + */ + readonly cluster: ICluster; + + /** + * The image to start. + */ + readonly image: ContainerImage; + + /** + * The CMD value to pass to the container. A string with commands delimited by commas. + * + * @default none + */ + readonly command?: string; + + /** + * Number of desired copies of running tasks + * + * @default 1 + */ + readonly desiredTaskCount?: number; + + /** + * Flag to indicate whether to enable logging + * + * @default true + */ + readonly enableLogging?: boolean; + + /** + * The environment variables to pass to the container. + * + * @default none + */ + readonly environment?: { [key: string]: string }; + + /** + * A name for the queue. + * + * If specified and this is a FIFO queue, must end in the string '.fifo'. + * + * @default CloudFormation-generated name + */ + readonly queueName?: string; + + /** + * Maximum capacity to scale to. + * + * @default (desiredTaskCount * 2) + */ + readonly maxScalingCapacity?: number + + /** + * The intervals for scaling based on the SQS queue's ApproximateNumberOfMessagesVisible metric. + * + * Maps a range of metric values to a particular scaling behavior. + * https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-scaling-simple-step.html + * + * @default [{ upper: 0, change: -1 },{ lower: 100, change: +1 },{ lower: 500, change: +5 }] + */ + readonly scalingSteps: autoScaling.ScalingInterval[]; +} +``` + +### Part 2: Create `Ec2QueueWorkerServiceProps` interface and `Ec2QueueWorkerService` construct + +The `Ec2QueueWorkerServiceProps` interface will contain properties to construct the Ec2TaskDefinition, SQSQueue and Ec2Service: + +```ts +/** + * Properties to define an ECS service + */ +export interface Ec2QueueWorkerServiceProps { + /** + * The minimum number of CPU units to reserve for the container. + * + * @default none + */ + readonly cpu?: number; + + /** + * The hard limit (in MiB) of memory to present to the container. + * + * If your container attempts to exceed the allocated memory, the container + * is terminated. + * + * At least one of memoryLimitMiB and memoryReservationMiB is required for non-Fargate services. + */ + readonly memoryLimitMiB?: number; + + /** + * The soft limit (in MiB) of memory to reserve for the container. + * + * When system memory is under contention, Docker attempts to keep the + * container memory within the limit. If the container requires more memory, + * it can consume up to the value specified by the Memory property or all of + * the available memory on the container instance—whichever comes first. + * + * At least one of memoryLimitMiB and memoryReservationMiB is required for non-Fargate services. + */ + readonly memoryReservationMiB?: number; +} +``` + +An example use case: +```ts +// Create the vpc and cluster used by the Queue Worker task +const vpc = new ec2.VpcNetwork(stack, 'Vpc', { maxAZs: 1 }); +const cluster = new ecs.Cluster(stack, 'EcsCluster', { vpc }); +cluster.addCapacity('DefaultAutoScalingGroup', { + instanceType: new ec2.InstanceType('t2.micro') +}); + +// Create the Queue Worker task +new Ec2QueueWorkerService(stack, 'EcsQueueWorkerService', { + cluster, + image: ecs.ContainerImage.fromRegistry('amazon/amazon-ecs-sample'), + desiredTaskCount: 2, + maxScalingCapacity: 5, + memoryReservationMiB: 512, + cpu: 256, + queueName: 'EcsWorkerQueue' +}); +``` + +### Part 3: Create `FargateQueueWorkerServiceProps` interface and `FargateQueueWorkerService` construct + +The `FargateQueueWorkerServiceProps` interface will contain properties to construct the FargateTaskDefinition, SQSQueue and FargateService: + +```ts +/** + * Properties to define an Fargate service + */ +export interface FargateQueueWorkerServiceProps { + /** + * The number of cpu units used by the task. + * Valid values, which determines your range of valid values for the memory parameter: + * 256 (.25 vCPU) - Available memory values: 0.5GB, 1GB, 2GB + * 512 (.5 vCPU) - Available memory values: 1GB, 2GB, 3GB, 4GB + * 1024 (1 vCPU) - Available memory values: 2GB, 3GB, 4GB, 5GB, 6GB, 7GB, 8GB + * 2048 (2 vCPU) - Available memory values: Between 4GB and 16GB in 1GB increments + * 4096 (4 vCPU) - Available memory values: Between 8GB and 30GB in 1GB increments + * + * This default is set in the underlying FargateTaskDefinition construct. + * + * @default 256 + */ + readonly cpu?: string; + + /** + * The amount (in MiB) of memory used by the task. + * + * This field is required and you must use one of the following values, which determines your range of valid values + * for the cpu parameter: + * + * 0.5GB, 1GB, 2GB - Available cpu values: 256 (.25 vCPU) + * + * 1GB, 2GB, 3GB, 4GB - Available cpu values: 512 (.5 vCPU) + * + * 2GB, 3GB, 4GB, 5GB, 6GB, 7GB, 8GB - Available cpu values: 1024 (1 vCPU) + * + * Between 4GB and 16GB in 1GB increments - Available cpu values: 2048 (2 vCPU) + * + * Between 8GB and 30GB in 1GB increments - Available cpu values: 4096 (4 vCPU) + * + * This default is set in the underlying FargateTaskDefinition construct. + * + * @default 512 + */ + readonly memoryMiB?: string; +} +``` + +An example use case: +```ts +// Create the vpc and cluster used by the Queue Worker task +const vpc = new ec2.VpcNetwork(stack, 'Vpc', { maxAZs: 2 }); +const cluster = new ecs.Cluster(stack, 'FargateCluster', { vpc }); + +// Create the Queue Worker task +new FargateQueueWorkerService(stack, 'FargateQueueWorkerService', { + cluster, + image: ecs.ContainerImage.fromRegistry('amazon/amazon-ecs-sample'), + desiredTaskCount: 2, + maxScalingCapacity: 5, + queueName: 'FargateWorkerQueue' +}); +```