httpcheck returns 1 even if http_status_class & http_status_code is not the same

[gfelot]

Component(s)

receiver/httpcheck

What happened?

Description

Getting status code from a healthcheck URL with receivers/httpcheck of services that is up return 1 also when http_status_class & http_status_code are not equal.

Steps to Reproduce

receivers:  
  httpcheck: # https://signoz.io/blog/health-check-monitoring-with-opentelemetry/
    collection_interval: 30s
    targets:
      - endpoint: "https://<ip_server>/healthcheck"
        method: GET
        tls:
          insecure_skip_verify: true
[...]
exporters:
  clickhousemetricswrite:
    endpoint: tcp://<signoz_serveur>:9000/signoz_metrics?username=u&password=p
    tls:
      insecure: true
    resource_to_telemetry_conversion:
      enabled: true

Expected Result

When service is up, http_status_code: 200 is sent so only the line with http_status_class: 2xx should be at one.

When I request in clickhouse (SigNoz stack) with

SELECT *
FROM signoz_metrics.time_series_v4
WHERE (metric_name = 'httpcheck_status') AND (JSONExtractString(labels, 'service_name') = 'PASSPORT') AND (JSONExtractString(labels, 'vm_hostname') = '3dx-dev08')
LIMIT 10

I got answer like

Row 3:
──────
env:            DEV
temporality:    Cumulative
metric_name:    httpcheck_status
description:    1 if the check resulted in status_code matching the status_class, otherwise 0.
unit:           1
type:           Sum
is_monotonic:   false
fingerprint:    2470872169406690154 -- 2.47 quintillion
unix_milli:     1745697600000 -- 1.75 trillion
labels:         {"__name__":"httpcheck_status","__temporality__":"Cumulative","deployment_environment":"DEV","health_score":"3","host_name":"FRPAR3DXDEV08","http_method":"GET","http_status_class":"5xx","http_status_code":"200","http_url":"https://<ip_serveur>/3dpassport/healthcheck","os_type":"linux","service_name":"PASSPORT","vm_hostname":"3dx-dev08"}
attrs:          {}
scope_attrs:    {}
resource_attrs: {}
__normalized:   true

I can see unit: 1 and I interprete this as 1.

For exemple in this 2xx check

Row 7:
──────
env:            DEV
temporality:    Cumulative
metric_name:    httpcheck_status
description:    1 if the check resulted in status_code matching the status_class, otherwise 0.
unit:           1
type:           Sum
is_monotonic:   false
fingerprint:    9321289193155664995 -- 9.32 quintillion
unix_milli:     1745697600000 -- 1.75 trillion
labels:         {"__name__":"httpcheck_status","__temporality__":"Cumulative","deployment_environment":"DEV","health_score":"3","host_name":"FRPAR3DXDEV08","http_method":"GET","http_status_class":"2xx","http_status_code":"200","http_url":"https://<ip_serveur>/3dpassport/healthcheck","os_type":"linux","service_name":"PASSPORT","vm_hostname":"3dx-dev08"}
attrs:          {}
scope_attrs:    {}
resource_attrs: {}
__normalized:   true

Actual Result

Getting 1 for 2xx, 0 for 3xx to 5xx.

Collector version

v0.111.39

Environment information

Environment

OS: RedHat 8.10
Compiler: Compilled with go version go1.23.6 (Red Hat 1.23.6-1.module+el8.10.0+22945+b2c96a17) linux/amd64

OpenTelemetry Collector configuration

Log output

Additional context

No response

[github-actions]

Pinging code owners:

• receiver/httpcheck: @codeboten @VenuEmmadi

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[AtiX]

Adding onto that, it seems like the metric for http_status_class=2xx and http_status_code=200 still stays at 1 even if the target host returns an error.

What you receive are two data points (e.g. when the service returns a 502 error):

• http_status_class=2xx and http_status_code=502 = 0 (expected, the status code does not match the status_class)
• http_status_class=2xx and http_status_code=200 = 1 (unexpected, the service is returning a 502, so the value should not be 1)