By using AWS re:Post, you agree to the AWS re:Post Terms of Use
AWS re:Postre:Post

AppRunner request timeout / HTTP 503

0

I've been successfully using App Runner to deploy a containerised microservice focused around uploading large files. The Docker container hosts a Gunicorn instance with a Flask web app.

I've that for larger file uploads (which naturally take longer), App Runner will answer with an HTTP 503 response. This happens after 30 seconds so I'm assuming there is a built-in request timeout enforced by App Runner, probably by the SSL-terminating proxy or load balancer.

Unfortunately, the App Runner documentation does not mention anything regarding request timeouts or maximum body sizes, so I can only guess. Before I change the upload to use a different mechanism (probably chunked upload), or move to a service with more control (Fargate?), I'd like to first confirm my suspicion. Also perhaps there is an undocumented way to work around this problem.

And yes, I increased the timeout of the application's HTTP server (gunicorn) to 600 seconds and also tested it locally, so there is definitely something inside App Runner causing this behavior.

Any help and/or pointers are greatly appreciated.

Topics
MicroservicesComputeContainers
Tags
MicroservicesAWS App Runner
Language
English
CUBE Project
asked 2 years ago1575 views
2 Answers
1
Accepted Answer

According to AWS support, this is a known limitation of App Runner (requests running longer than 30 seconds result in HTTP 503), and is scheduled to be addressed in a future update. No ETA given unfortunately.

CUBE Project
answered 2 years ago
  • Jamie Visker
    2 years ago

    Is this documented somewhere?

  • Patrick
    4 months ago

    The timeout per request is currently 120 sec.

0

I don't think we cut hard the connectivity but we do throttle the CPU down when there is no active requests to the instance. See this thread.

AWS
EXPERT
mreferre
answered 2 years ago
  • mreferre EXPERT
    2 years ago

    How/where do you issue the curl command? Do you schedule it from the outside?

  • CUBE Project
    2 years ago

    Hi. I issue the curl command from outside, from a normal internet-connected host, outside of AWS. For me this strongly feels like a 30 second timeout enforced by either load balancer or envoy proxy.

  • CUBE Project
    2 years ago

    I investigated this further and added a route in the micro service that just sleeps for a specified amount of seconds and then returns an HTTP 200 response, to eliminate any other factors in the application.

    $ curl -v https://.../sleep?seconds=31

    GET /sleep?seconds=31 HTTP/1.1 Host: <redcated>.eu-west-1.awsapprunner.com User-Agent: curl/7.64.1 Accept: /

    < HTTP/1.1 503 Service Unavailable < content-length: 95 < content-type: text/plain < date: Wed, 29 Dec 2021 13:19:47 GMT < server: envoy < connection: close

    Sleeping for a shorter time (e.g. 29 seconds or less) works fine.

  • CUBE Project
    2 years ago

    I will try approaching AWS support with this. Thanks for the pointer though!

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions

Relevant content