Fails to update existing ECS non-loadbalanced service when adding to load balancer

Describe the bug

When there is an existing ECS service which is not initially configured with a load balancer, and a change is made to add this service to a target group for an ALB, the CFN change performed is an UPDATE. However this is not possible since ALB assignment can only be done to an ECS service during creation.

Expected Behavior

In this scenario the change should recreate the service resource rather than update it.

Current Behavior

Update fails with the following message:

Resource handler returned message: "Invalid request provided: UpdateService error: Health check grace period is only valid for services configured to use load balancers (Service: AmazonECS; Status Code: 400; Error Code: InvalidParameterException; Request ID: xxxyyyzzz; Proxy: null)"

Reproduction Steps

1. Create a FargateService which is not backed by a load balancer.
2. Deploy
3. Add the service from step 1 as a target to an ALB’s listener
4. Deploy

Possible Solution

Rather than making the change as an update, it should recreate the service resource in order to assign the ALB correctly.

Additional Information/Context

No response

CDK CLI Version

2.20.0 (build 738ef49)

Framework Version

No response

Node.js Version

v17.8.0

OS

Ubuntu 21.10 (5.17.0-051700-generic)

Language

Typescript

Language Version

4.6.2

Other information

No response

[jaecktec]

(service.node.defaultChild as CfnService).healthCheckGracePeriodSeconds = undefined;

works for me

[peterwoodworth]

I’m not sure how I ran into the deployment stall after i deleted the health check from my template. Must have been an intermittent issue 😕

We should be able to support adding a load balancer to a service without requiring that users use an escape hatch to remove the property