GitHub receiver step spans show huge times when step is skipped

[adrielp]

Component(s)

receiver/github

What happened?

Description

The step spans for skipped runs show times consistently show times like: 4639920h 46m. I'm mainly opening this up for myself to track and am actively looking at the fix. Wanted to provide transparency though on the discovery.

Steps to Reproduce

Any job with skipped steps will show this way.

Expected Result

Expected result is 0 time w/ the status of skipped.

Actual Result

Actual result is a crazy huge time w/ a status of skipped as depicted below.

Collector version

v0.122.0

Environment information

Environment

All

OpenTelemetry Collector configuration

Log output

Additional context

No response

[github-actions]

Pinging code owners:

• receiver/github: @adrielp @andrzej-stencel @crobert-1 @TylerHelmuth

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

[bacherfl]

Hi @adrielp! Thank you for creating the issue - may i ask to also include the collector config, so it's also easier to reproduce for others?

[arianvp]

Reproduces with the following config:

receivers:
  github:
    initial_delay: 1s
    collection_interval: 60s
    scrapers:
      scraper:
        github_org: MercuryTechnologies
        auth:
          authenticator: bearertokenauth/github
    webhook:
      secret: "${env:GITHUB_WEBHOOK_SECRET}"
      endpoint: 0.0.0.0:8080
      path: /events
      health_path: /health

processors:
  batch:
    timeout: 10s
    send_batch_size: 1024

exporters:
  otlp:
    endpoint: "api.honeycomb.io:443"
    headers:
      "x-honeycomb-team": "${env:HONEYCOMB_API_KEY}"
      "x-honeycomb-dataset": "${env:HONEYCOMB_DATASET}"
    tls:
      insecure: false
  debug: # To use locally for testing/debugging
    verbosity: detailed
    sampling_initial: 5
    sampling_thereafter: 200
    
extensions:
  bearertokenauth/github:
    token: "${env:GITHUB_PAT}"
  health_check:
    endpoint: 0.0.0.0:13133

service:
  extensions: [health_check, bearertokenauth/github]
  pipelines:
    traces:
      receivers:
        - github
      processors:
        - batch
      exporters:
        - otlp

[adrielp]

Yea, it's the default behavior for the config w/ any traces. GitHub does a weird thing when emitting skipped jobs. Soon as I'm back from Kubecon I'll have a fix up for this. Largely wanted to open this up for myself to go fix (though if someone wants to take it on, no objections).

@arianvp thanks for providing the config.

[arianvp]

Enjoy KubeCon!

[arianvp]

Also happens for cancelled jobs

[adrielp]

Discussed in the CICD SIG today. Requires fix.

[adrielp]

Pull request to fix this issue is open.

[arianvp]

This issue is not fixed for me

Maybe some interation with #39511 ? it seems to happen to the queued span

[adrielp]

@arianvp -- can you show me what the start and stop time is on those spans? -- please feel free to troubleshoot with me more quickly on slack.