Alternative for setting request timeouts for in-process hosting

[guardrex]

From @effyteva on #5449 ...

As "ANCM: RequestTimeout doesn't apply to inProcess aspnet/IISIntegration#555",

What's the recommended alternative for setting request timeouts when using in process?

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: fe69b3a9-896f-e189-0c2c-f87ff735df9c
• Version Independent ID: 97c8549b-66ad-0a23-1971-784fb923042a
• Content: ASP.NET Core Module
• Content Source: aspnetcore/host-and-deploy/aspnet-core-module.md
• Product: aspnet-core
• Technology: aspnetcore-hostdeploy
• GitHub Login: @guardrex
• Microsoft Alias: riande

[effyteva]

Thank you @guardrex
I've actually kept digging, and figured out it's even a bit harder to work with than I thought.
I tried switching to OutOfProcess, in which the RequestTimeout parameter works.
The request was aborted and the client got the error page "The specified CGI application encountered an error and the server terminated the process.".

However, the server keeps processing infinitely, which left the server in an invalid state, until it is restarted.

Re-producing is very easy, I created a Razor View, with a "while (true)" loop in it.
I couldn't get it to abort after any kind of timeout setting.

[guardrex]

Yes, and your experience matches what engineering said on that issue ...

The in-proc corollary is waiting for the application to process the request.

Since any app can just swallow the request and we don't address that scenario anywhere in the docs as far as I know, you ask a great question. What we need to do on this issue is mark it for the Product Unit ("PU" label) so that they can tell us what they think about an alternative.

Now that Preview4 is out, I think (hope) that they'll look at this soon. I'll take their guidance and enhance that remark in the topic. We should at the very least say what engineering said: The app is left to process the request without a timeout by IIS/ANCM.

Note to shirhatti: I'll work this and remove the "PU" label when there's guidance available.

[effyteva]

Thanks, I highly appreciate your quick and detailed responses.
Let's see what the Product Units team comes up with, hopefully there's a simple and effective way to abort these abandoned requests.

[guardrex]

This is old but interesting ... might lead to some newer info or an approach.

aspnet/KestrelHttpServer#611

[effyteva]

This is old but interesting ... might lead to some newer info or an approach.

aspnet/KestrelHttpServer#611

Sadly it's related to ASP.NET and not ASP.NET Core.
The Server property doesn't exist anymore.

[guardrex]

Kestrel is the server in that scenario, so they're discussing how it would work with Core.

[guardrex]

Ben has an "execution timeout" ... I'll track that down and see if it went in.

[guardrex]

Doesn't look like it ... https://docs.microsoft.com/dotnet/api/microsoft.aspnetcore.server.kestrel.core.kestrelserverlimits

[guardrex]

... and here's why ... aspnet/KestrelHttpServer#485 (comment)

[effyteva]

I see, not sure how to proceed from here,
I understand the issue with Tasks not being able to Abort ("kill") the same way Threads could've been aborted.
However, this means the in case of a CPU bound loop (such as the "while (true)"), the only solution is to restart Kestrel completely, which doesn't make much sense.