500 Errors showing as 200 in /metrics on ASP.NET Core 2.2

I have an ASP.NET Core 2.2 project with one controller that works like this:

• Get() method = 200
• Post() method = 500 (throws exception)
• Put() method = 400 (returns bad request)

When testing I am confirming the Post method returns a 500 via CURL, but in /metrics it shows up as a 200 for the ValuesController Post method.

Output of /metrics:

http_request_duration_seconds_sum{code="200",method="POST",controller="Values",action="Post"} 0.1010758
http_request_duration_seconds_count{code="200",method="POST",controller="Values",action="Post"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.001"} 0
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.002"} 2
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.004"} 2
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.008"} 2
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.016"} 2
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.032"} 2
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.064"} 2
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.128"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.256"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="0.512"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="1.024"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="2.048"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="4.096"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="8.192"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="16.384"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="32.768"} 3
http_request_duration_seconds_bucket{code="200",method="POST",controller="Values",action="Post",le="+Inf"} 3

Controller:

// GET api/values
        [HttpGet()]
        public ActionResult Get()
        {
            Ok();
        }

        [HttpPost()]
        public ActionResult<IEnumerable<string>> Post(string text)
        {
            _exceptionCount++;
            throw new Exception($"I am exception number {_exceptionCount}");
        }

        [HttpPut()]
        public ActionResult Put()
        {
            return BadRequest();
        }

Curl example shows 500 error:

C02XX041JGH7:~ lin.meyer$ echo '{"text": "Hello **world**!"}' | curl -X POST -d @- --verbose http://localhost:5000/api/values
Note: Unnecessary use of -X or --request, POST is already inferred.
*   Trying ::1...
* TCP_NODELAY set
* Connected to localhost (::1) port 5000 (#0)
> POST /api/values HTTP/1.1
> Host: localhost:5000
> User-Agent: curl/7.54.0
> Accept: */*
> Content-Length: 28
> Content-Type: application/x-www-form-urlencoded
>
* upload completely sent off: 28 out of 28 bytes
< HTTP/1.1 500 Internal Server Error
< Date: Mon, 15 Jul 2019 17:47:06 GMT
< Server: Kestrel
< Content-Length: 0
<
* Connection #0 to host localhost left intact

Startup:

        public void Configure(IApplicationBuilder app, IHostingEnvironment env)
        {
            app.UseHttpMetrics(); // Adds basic HTTP metrics to Prometheus /metrics (must be before UseMvc)
            app.UseMetricServer();
            app.UseHealthChecks("/health");
            app.UseMvc();
        }

[mrich316]

I always use the ExceptionHandlerMiddleware after prometheus-net and I don’t see this behaviour

app.UseMetricServer("/diagnostics/v1/metrics");
app.UseHttpMetrics();
app.UseExceptionHandler(...);

And prometheus-net seems to report the proper thing. Do I miss something ?

[sandersaares]

If you use an explicit exception handler middleware and configure it after the metrics middleware, that’s fine. However, the problem occurs when you either configure the exception handler before metrics or when you do not configure one at all (and let ASP.NET fallback to its default logic, which emits a 500).

After doing some digging in the ASP.NET Core source, I have not found a meaningful way to improve the behavior here, so I will fall back to documenting this aspect. In the end, ASP.NET Core is a pipeline and the nature of the pipelined processing must be understood for actual behavior to match expected behavior. This situation is one of the sufficiently nontrivial ones where it starts to matter.