Metric - Error in collected metric http_server_duration

[julianocosta89]

Bug Report

Which version of the demo you are using?
6af2bdc.

Symptom

A clear and concise description of what the bug is.

What is the expected behavior?
I expect the sample to run without any errors.

What is the actual behavior?

After starting the sample, I started getting the following error on the OTel collector logs:

otel-col                 | * collected metric http_server_duration label:<name:"container_id" value:"354b4b881bf8956b0483f2231f348cd99f4a56e087617808e1ff88ac435400d2" > label:<name:"http_flavor" value:"HTTP/2" > label:<name:"http_host" value:"cartservice:7070" > label:<name:"http_method" value:"POST" > label:<name:"http_scheme" value:"http" > label:<name:"http_status_code" value:"200" > label:<name:"http_target" value:"/hipstershop.CartService/GetCart" > label:<name:"job" value:"opentelemetry-demo/cartservice" > label:<name:"service_name" value:"cartservice" > label:<name:"service_namespace" value:"opentelemetry-demo" > label:<name:"telemetry_sdk_language" value:"dotnet" > label:<name:"telemetry_sdk_name" value:"opentelemetry" > label:<name:"telemetry_sdk_version" value:"1.4.0.665" > histogram:<sample_count:615 sample_sum:689.5118000000004 bucket:<cumulative_count:0 upper_bound:0 > bucket:<cumulative_count:613 upper_bound:5 > bucket:<cumulative_count:614 upper_bound:10 > bucket:<cumulative_count:614 upper_bound:25 > bucket:<cumulative_count:614 upper_bound:50 > bucket:<cumulative_count:614 upper_bound:75 > bucket:<cumulative_count:614 upper_bound:100 > bucket:<cumulative_count:615 upper_bound:250 > bucket:<cumulative_count:615 upper_bound:500 > bucket:<cumulative_count:615 upper_bound:750 > bucket:<cumulative_count:615 upper_bound:1000 > bucket:<cumulative_count:615 upper_bound:2500 > bucket:<cumulative_count:615 upper_bound:5000 > bucket:<cumulative_count:615 upper_bound:7500 > bucket:<cumulative_count:615 upper_bound:10000 > >  has help "measures the duration of the inbound HTTP request" but should have "measures the duration of the inbound HTTP requests"
otel-col                 |      {"kind": "exporter", "data_type": "metrics", "name": "prometheus"}
otel-col                 | github.com/open-telemetry/opentelemetry-collector-contrib/exporter/prometheusexporter.(*promLogger).Println
otel-col                 |      github.com/open-telemetry/opentelemetry-collector-contrib/exporter/prometheusexporter@v0.70.0/log.go:34
otel-col                 | github.com/prometheus/client_golang/prometheus/promhttp.HandlerForTransactional.func1
otel-col                 |      github.com/prometheus/client_golang@v1.14.0/prometheus/promhttp/http.go:139
otel-col                 | net/http.HandlerFunc.ServeHTTP
otel-col                 |      net/http/server.go:2109
otel-col                 | net/http.(*ServeMux).ServeHTTP
otel-col                 |      net/http/server.go:2487
otel-col                 | go.opentelemetry.io/collector/config/confighttp.(*decompressor).wrap.func1
otel-col                 |      go.opentelemetry.io/collector@v0.70.0/config/confighttp/compression.go:162
otel-col                 | net/http.HandlerFunc.ServeHTTP
otel-col                 |      net/http/server.go:2109
otel-col                 | go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp.(*Handler).ServeHTTP
otel-col                 |      go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp@v0.37.0/handler.go:210
otel-col                 | go.opentelemetry.io/collector/config/confighttp.(*clientInfoHandler).ServeHTTP
otel-col                 |      go.opentelemetry.io/collector@v0.70.0/config/confighttp/clientinfohandler.go:39
otel-col                 | net/http.serverHandler.ServeHTTP
otel-col                 |      net/http/server.go:2947
otel-col                 | net/http.(*conn).serve
otel-col                 |      net/http/server.go:1991

Reproduce

Clone the repo and simply run:

docker compose up

We will close this issue if:

• The steps you provided are complex.
• If we can not reproduce the behavior you're reporting.

Additional Context

The app runs normally, but from the logs it seems that some metrics are not being sent to Prometheus.

[austinlparker]

Hm, this seems like a mismatch between metrics somehow. Do we have the same metric in two places with mismatched help?

[austinlparker]

Interesting, that metric seems to have changed. At head...

# TYPE http_server_duration histogram
http_server_duration_bucket{http_flavor="HTTP/2",http_host="cartservice:7070",http_method="POST",http_scheme="http",http_status_code="200",http_target="/hipstershop.CartService/AddItem",job="opentelemetry-demo/cartservice",service_name="cartservice",service_namespace="opentelemetry-demo",telemetry_sdk_language="dotnet",telemetry_sdk_name="opentelemetry",telemetry_sdk_version="1.4.0.665",le="0"} 0

1.3.0

# HELP http_server_duration measures the duration of the inbound HTTP requests
# TYPE http_server_duration histogram
http_server_duration_bucket{container_id="8d004e7bce6343dfc65a4f5d7a562750588b7966e8ed4b05275e03aaac1c760d",host_arch="arm64",host_name="8d004e7bce63",http_flavor="1.1",http_method="GET",http_scheme="http",http_status_code="200",job="opentelemetry-demo/frontend",net_host_name="frontend",net_host_port="8080",os_type="linux",os_version="5.15.49-linuxkit",process_command="/app/server.js",process_command_line="/usr/local/bin/node /app/server.js",process_executable_name="node",process_pid="17",process_runtime_description="Node.js",process_runtime_name="nodejs",process_runtime_version="18.14.0",service_name="frontend",service_namespace="opentelemetry-demo",le="0"} 0

However, I think what's happening is that these metrics are in head and 1.3, but what's happening is that they're racing each other. The problem is, the #HELP text differs between them (which I believe is the OTLP metric description), which causes the prometheus exporter to not merge them correctly, but whichever one gets scraped + reported first is the one that 'wins' and prometheus won't overwrite what's already there?

[austinlparker]

Keeping this open as we'll need to take the release from .NET which fixes it upstream.

[puckpuck]

I tried taking the latest 1.4.0/1.0.0-rc9.14 from the .NET SDKs and still got the error. This PR should fix it. I will test it again after it's released.

[cartersocha]

Looks like we need to manually request a release. open-telemetry/opentelemetry-dotnet-contrib#1063 (comment)

@puckpuck would our ask be to update this to rc9.9?

https://github.com/open-telemetry/opentelemetry-dotnet-contrib/releases/tag/Instrumentation.AspNet.TelemetryHttpModule-1.0.0-rc9.8

[puckpuck]

I worked with @martinjt on this ... and it turns out I was barking up the wrong tree and I misunderstood how .NET instrumentation packages were done 🤦

We still see this error because JS also had the same incorrect description that the .NET SDK had. I have this PR in the JS repo to address this.

log entry from collector

otel-col  | * collected metric http_server_duration label:<name:"container_id" value:"e7414d44db0e8b8029644b59a47717a325a6c420f62dc08b293dc30c4b750c71" > label:<name:"host_arch" value:"arm64" > label:<name:"host_name" value:"e7414d44db0e" > label:<name:"http_flavor" value:"1.1" > label:<name:"http_method" value:"GET" > label:<name:"http_scheme" value:"http" > label:<name:"http_status_code" value:"200" > label:<name:"job" value:"opentelemetry-demo/frontend" > label:<name:"net_host_name" value:"frontend" > label:<name:"net_host_port" value:"8080" > label:<name:"os_type" value:"linux" > label:<name:"os_version" value:"5.15.49-linuxkit" > label:<name:"process_command" value:"/app/server.js" > label:<name:"process_command_line" value:"/usr/local/bin/node /app/server.js" > label:<name:"process_executable_name" value:"node" > label:<name:"process_pid" value:"16" > label:<name:"process_runtime_description" value:"Node.js" > label:<name:"process_runtime_name" value:"nodejs" > label:<name:"process_runtime_version" value:"18.15.0" > label:<name:"service_name" value:"frontend" > label:<name:"service_namespace" value:"opentelemetry-demo" > label:<name:"telemetry_sdk_language" value:"nodejs" > label:<name:"telemetry_sdk_name" value:"opentelemetry" > label:<name:"telemetry_sdk_version" value:"1.10.1" > histogram:<sample_count:122 sample_sum:3952 bucket:<cumulative_count:0 upper_bound:0 > bucket:<cumulative_count:19 upper_bound:5 > bucket:<cumulative_count:46 upper_bound:10 > bucket:<cumulative_count:103 upper_bound:25 > bucket:<cumulative_count:115 upper_bound:50 > bucket:<cumulative_count:116 upper_bound:75 > bucket:<cumulative_count:116 upper_bound:100 > bucket:<cumulative_count:118 upper_bound:250 > bucket:<cumulative_count:118 upper_bound:500 > bucket:<cumulative_count:122 upper_bound:1000 > >  has help "measures the duration of the inbound HTTP requests" but should have "Measures the duration of inbound HTTP requests."

[austinlparker]

The JS PR to fix this was merged (open-telemetry/opentelemetry-js#3710)

[julianocosta89]

we just need a new release now 🤞🏽

[puckpuck]

I believe this is all resolved now, and we can close this issue.