ApplicationLoadBalancedFargateService with listener on one port and health check on another fails health check

Question

Hi, I have an ApplicationLoadBalancedFargateService that exposes a service on one port, but the health check runs on another.  Unfortunately, the target fails health check and terminates the task.  Here's a snippet of my code

```
    const hostPort = 5701;
    const healthCheckPort = 8080;

taskDefinition.addContainer(stackPrefix + 'Container', {
      image: ecs.ContainerImage.fromRegistry('hazelcast/hazelcast:3.12.6'),
      environment : {
        'JAVA_OPTS': `-Dhazelcast.local.publicAddress=localhost:${hostPort} -Dhazelcast.rest.enabled=true`, 
        'LOGGING_LEVEL':'DEBUG',
        'PROMETHEUS_PORT': `${healthCheckPort}`},
        portMappings: [{containerPort : hostPort, hostPort: hostPort},{containerPort : healthCheckPort, hostPort: healthCheckPort}],
        logging: ecs.LogDriver.awsLogs({streamPrefix: stackPrefix, logRetention: logs.RetentionDays.ONE_DAY}),
    });

const loadBalancedFargateService = new ecsPatterns.ApplicationLoadBalancedFargateService(this, stackPrefix + 'Service', {
      cluster,
      publicLoadBalancer : false,
      desiredCount: 1,
      listenerPort: hostPort,
      taskDefinition: taskDefinition,
      securityGroups : [fargateServiceSecurityGroup],
      domainName : env.getPrefixedRoute53(stackName),
      domainZone : env.getDomainZone(),
    });

loadBalancedFargateService.targetGroup.configureHealthCheck({
      path: "/metrics", 
      port: healthCheckPort.toString(),
      timeout: cdk.Duration.seconds(15),  
      interval: cdk.Duration.seconds(30),
      healthyThresholdCount: 2,
      unhealthyThresholdCount: 5,
      healthyHttpCodes: '200-299'
    });

```
Any suggestions on how I can get this to work?  thanks

Answer

Hello there,

Looking into your code, a few things I noticed:

1. For Fagete Service, there is no need to define host port, since the host port and the container port should be the same.

2. You should use `taskImageOptions` tp define container port

Here is an example code snippet:

```
    const service = new EcsPatterns.ApplicationLoadBalancedFargateService(this, 'demoService', {
      cluster: cluster,
      publicLoadBalancer: true,
      certificate,
      memoryLimitMiB: 2048,
      cpu: 512,
      desiredCount: 2,
      serviceName: 'demoService',
      taskImageOptions: {
        containerName: 'web',
        containerPort: 8080,   // Your container or app port
        executionRole,
        taskRole,
        environment: {
          ...
        },
        image: ecs.ContainerImage.fromEcrRepository(repository),
        logDriver: new ecs.AwsLogDriver({
          logGroup,
          streamPrefix: props.serviceName,
        }),
      },
    })

service.targetGroup.configureHealthCheck({
      path: '/v1/healthcheck',
    })

```

With above example, `ApplicationLoadBalancedFargateService` configures health check against container port 8080, but with a different path `/v1/healthcheck` as per overriding.

ApplicationLoadBalancedFargateService with listener on one port and health check on another fails health check

Relevant questions

ApplicationLoadBalancedFargateService with listener on one port and health check on another fails health check

ApplicationLoadBalancedFargateService with load balancer, target groups, targets on non-standard port

Health check at NLB level for a Fargate Service

LoadBalancer health check fails but instance is not terminating

Health check on custom port number fails.

Health check in route53 by port

NLB-ECS Health Check

Stopping health check

Apprunner fails on Health check

identical nlbs, one has intermittent health check failures