Why can't I run my ECS task from AWS Lambda? - amazon-web-services

Question

I am using Amazon Web Services and trying to run an ECS Task Definition on a Cluster triggered from a Lambda.
When I run this task manually in the ECS console and chose all of the same options as I'm passing to run_task, it runs just fine. I see logs in Cloudwatch and the effects of the task (updaing a database) have happened as expected. But when I run the task from a Lambda it does not work, but also gives me no errors that I can see.
Here's the Lambda definition:
import boto3
def lambda_handler(event, context):
print("howMuchSnowDoUpdate")
client = boto3.client('ecs')
response = client.run_task(
cluster='HowMuchSnow',
taskDefinition='HowMuchSnow:2',
count=1,
launchType='FARGATE',
networkConfiguration={
'awsvpcConfiguration': {
'subnets': [
'subnet-ebce7c8c',
],
'securityGroups': [
'sg-03bb63bf7b3389d42',
],
'assignPublicIp': 'DISABLED'
}
},
)
print(response)
I have given the Lambda's IAM role the policy of ECSFull. Before I did I was getting an expected permission denied when running run_task. But once I added that policy, the Lambda runs just fine with no errors reported and this is the response that I get from that print(response) line:
{'tasks': [{'taskArn': 'arn:aws:ecs:us-east-1:221691463461:task/10b2473f-482d-4f75-ab43-3980f6995b17', 'clusterArn': 'arn:aws:ecs:us-east-1:221691463461:cluster/HowMuchSnow', 'taskDefinitionArn': 'arn:aws:ecs:us-east-1:221691463461:task-definition/HowMuchSnow:2', 'overrides': {'containerOverrides': [{'name': 'HowMuchSnow'}]}, 'lastStatus': 'PROVISIONING', 'desiredStatus': 'RUNNING', 'cpu': '256', 'memory': '512', 'containers': [{'containerArn': 'arn:aws:ecs:us-east-1:221691463461:container/9a76562b-1fef-457f-ae04-0f0eb4003e7b', 'taskArn': 'arn:aws:ecs:us-east-1:221691463461:task/10b2473f-482d-4f75-ab43-3980f6995b17', 'name': 'HowMuchSnow', 'lastStatus': 'PENDING', 'networkInterfaces': []}], 'version': 1, 'createdAt': datetime.datetime(2019, 6, 17, 14, 57, 29, 831000, tzinfo=tzlocal()), 'group': 'family:HowMuchSnow', 'launchType': 'FARGATE', 'platformVersion': '1.3.0', 'attachments': [{'id': 'e6ec4941-9e91-47d1-adff-d406f28b1931', 'type': 'ElasticNetworkInterface', 'status': 'PRECREATED', 'details': [{'name': 'subnetId', 'value': 'subnet-ebce7c8c'}]}]}], 'failures': [], 'ResponseMetadata': {'RequestId': '3a2506ef-9110-11e9-b57a-d7e334b6f5f7', 'HTTPStatusCode': 200, 'HTTPHeaders': {'x-amzn-requestid': '3a2506ef-9110-11e9-b57a-d7e334b6f5f7', 'content-type': 'application/x-amz-json-1.1', 'content-length': '1026', 'date': 'Mon, 17 Jun 2019 14:57:29 GMT'}, 'RetryAttempts': 0}}
To my eyes this looks alright. But the task never actually runs. I do see a pending task in tasks list in the ECS console for my cluster briefly. But it runs not nearly as long as the actual task should run. It produces no logs in CloudWatch like it does when I run manually. I see no errors in the logs either.
One thing I will note is that I have to pick a VPC when running the task manually from the console but that's not a valid argument to boto3's ECS run_task function so I don't pass it.
Anyone know what might be going wrong or where I might look for information?

Answer 1

Here's what works for me.
When setting up Lambda:
Role must have ECS run task abilities
Don't specify a VPC in the Lambda function settings itself
Here's the Lambda code (replacing subnets, security groups, etc. for your own).
import boto3
client = boto3.client('ecs')
cluster_name = "demo-cluster"
task_definition = "demo-task:1"
def lambda_handler(event, context):
try:
response = client.run_task(
cluster=cluster_name,
launchType = 'FARGATE',
taskDefinition=task_definition,
count = 1,
platformVersion='LATEST',
networkConfiguration={
'awsvpcConfiguration': {
'subnets': [
'subnet-0r6gh701',
'subnet-a73d7c10'
],
'securityGroups': [
"sg-54cb123f",
],
'assignPublicIp': 'ENABLED'
}
})
print(response)
return {
'statusCode': 200,
'body': "OK"
}
except Exception as e:
print(e)
return {
'statusCode': 500,
'body': str(e)
}

Answer 2

I had this problem and it turned out that I had commented out the CMD line at the end of my Dockerfile during debugging. As such the lambda ran, but no ECS task was logged. Uncommenting the CMD led to the ECS task running and logging again.