This post will talk about the “Fail to forward request” messages for your application on App Service Linux - what it means, and why you may see it.
The full message being discussed here is this one (the below is an example):
Failed to forward request to http://172.16.7.3:8080. Encountered a System.Threading.Tasks.TaskCanceledException exception after 300067.221ms with message: The operation was canceled.. Check application logs to verify the application is properly handling HTTP traffic.
NOTE: The IP and port in your message may differ
This can be seen in a number of places, notably Log Stream, Application Logs (Diagnose and Solve detector), and in the log file with the syntax of
This message will appear only when the
middleware container is running. You can read more about the other types of platform containers here - Why you may see Docker Containers that are not yours on App Service Linux.
The “middleware” container will activate if any of these services are activated - all three of these services are ran through here:
- AutoHeal (Linux)
- CORS (Enabled through the portal only)
- EasyAuth (Authentication blade)
docker.log, you’ll see this running with the
_middleware suffix -
myfakesite_2_abc12345_middleware. This container will generate its own log with the naming scheme of
IMPORTANT: Regardless of whether just CORS, EasyAuth, or AutoHeal is enabled - they will all create and write to
YYYY_MM_DD_machinename_easyauth_docker.log - this is because they all run through the
middleware container. This can appear misleading at times.
Why does this appear?
As the message states, the
middleware container was not able to forward the request to the application container successfully. This is usually the case due to an application issue/error. When the request is forwarded, if an error is encountered, the middleware logic will return this message indicating an issue has occurred to the client.
In a typical flow when this middleware container is involved, it forwards the request much like a proxy to the application.
Most times, but not always - an error (in the form of stderr from the application) should be logged out to
default_docker.log in addition to this.
An example of trying to reproduce this can be done like this - say we implement a
setTimeout() to only return a response after 240 seconds. On Azure, idle HTTP requests time out at this mark - so this is mimicking an upstream response on this route for the application taking an extended amount of time to respond, or, never returning a response - which will give us a HTTP 504.
setTimeout(() => console.log("Waiting for 240 seconds"), 240000);
We’ll ultimate see the below, this example was viewing logs through Diagnose and Solve Problems -> Application Logs
A note about the IP and Port in the message:
This IP and Port should be the IP and Port combination of the application container. A quick way to validate this is to SSH into the container, which will have the Container IP in the bottom left.
The port will be the port of the exposed port of the Docker Container. For Blessed Images, you can refer to this - Default exposed ports of Azure App Service Linux Blessed Images - and correlate this with the stack that is being used.
For custom Docker Images, you would have to review which port is being exposed by your custom container.
NOTE: For applications listening on port 80 - there will be no port in the URL. This is normal and expected
If this occurs, review these points to narrow down the application error:
- Ensure application logging is enabled and reviewed.
- Review for any dependencies on the route(s) being invoked as well as what logic is on these routes.
- Review the status code associated with errors around this time.
- Sometimes it’s warranted to instrument better logging through code - or - through testing, to see if a better error surfaces by disabling the services being used (Autoheal, CORS, or EasyAuth)
You can view logging in a few different ways:
- Diagnose and Solve Problems -> Application Logs (detector), and others
- Directly via the Kudu site and browsing Log Files (/home/LogFiles) - these can be downloaded as well
- Through an FTP client to view Log Files (/home/LogFiles)
- Application Insights or Log Analytic workspaces
Since this generally occurs due to an application-layer issue, some scenarios may be:
- Routes that invoke dependencies (external API’s, databases, etc.) are timing out (like in our above example) and not returning a response within the 240 second limit
- There is a runtime error - such as bad syntax/bad request, invalid request for the application, or other general HTTP 500-related errors.
- The application is running high on memory or CPU - if this is affecting the host, this can have a variety of issues, one being where requests do not successfully complete, especially if expensive application logic is being invoked.
Given that two of these 3 services that make the middleware container run require configuration for your application - i.e, CORS (portal) and/or EasyAuth - it would be good to review settings.
The built-in Authentication and Authorization documentation can be found here