Hello, I am trying to run an ECS task while defining the task definition as a dictionary (https://github.com/anna-geller/packaging-prefect-flows/blob/master/flows/s3_ecs_run_task_definition_as_dict.py), but I keep receiving this error:

An error occurred (ClientException) when calling the RegisterTaskDefinition operation: Container.image should not be null or empty

I'm have defined the task as a yaml file (e.g. https://github.com/anna-geller/packaging-prefect-flows/blob/master/flows/s3_ecs_run_custom_task_definition.py) and that works fine, so I am not sure what I am doing wrong. Thanks!

Yeah that’s a bit weird. This errors during runtime right? Is there something on the agent side like a task definition?

Yes the error is during runtime. I am trying to define the task definition within the task, but am going to test defining the task definition on the agent as well. I'm not sure what you mean my changing the name. Here is my run config:

RUN_CONFIG = ECSRun(
    labels=["prod"],
    task_definition=dict(
        family="*********",
        requiresCompatibilities=["FARGATE"],
        executionRoleArn="arn:aws:iam::*******:role/**********",
        taskRoleArn="arn:aws:iam::*********:role/*****",
        networkMode="awsvpc",
        cpu=512,
        memory=1024,
        containerDefinitions=[
            dict(
                name="*********",
                image="*****.<http://dkr.ecr.us-east-1.amazonaws.com/****:latest|dkr.ecr.us-east-1.amazonaws.com/****:latest>",
                essential=True,
                logConfiguration=
                    dict(
                        logDriver="awslogs",
                        options=
                            {
                                "awslogs-group":"/ecs/***",
                                "awslogs-region":"us-east-1",
                                "awslogs-stream-prefix":"ecs",
                            },
                    ),
            
                secrets=[
                    dict(
                        valueFrom="arn:aws:secretsmanager:*****",
                        name="secret_name",
                    )
                ],
            ),
        ],
    ),
    run_task_kwargs=dict(cluster="prefectEcsCluster"),
)

Prefect requires the container

name

to be named

flow

for the one that actually runs the Flow

Oh I see, that did the trick. Thanks!

No problem!

Hey @Kevin Kho, one additional question. Our Data Pipelines run via ECS Tasks. So that's what I am triggering above. The pipeline has run successfully, but it appears that flow is not recognizing it as successful. Instead, a Lazarus process is retrying the task. Is there something special that I need to return in the task so that prefect moves on to the next step?

You shouldn’t have to. Do all the tasks complete successfully?

Yes they do

That’s weird. Do you have an unclosed connection by chance?

I don't believe so, all the tasks were successfully stopped

Do you have a state handler at the end of the Flow? Is this happening all the time?

Yeah, happening each time - I have not added a state handler yet. I am just testing out some basic pieces first

Can you DM me your Flow so I can take a look? First time I’ve seen this behavior on ECS