Create "Job Inputs" Dashboard for Interpolation Metrics

Overview

This issue proposes creating a dedicated dashboard to visualize the Job Inputs interpolation metrics that were recently instrumented in gitlab-runner. The metrics track interpolation usage, failures, and error types across runner instances.

Metrics Being Visualized

The dashboard will visualize these Prometheus metrics introduced in !6047 (merged):

• gitlab_runner_job_inputs_interpolations_total: Total interpolations where expressions were actually used
• gitlab_runner_job_inputs_interpolation_failures_total: Failed interpolations, labeled by error type (parse error, evaluation error, unsupported sensitive values)

Proposed Solutions

Option 1 (Recommended): Create a new dedicated dashboard

• Location: job-inputs.dashboard.jsonnet (https://gitlab.com/gitlab-com/runbooks/-/blob/master/dashboards/ci-runners/job-inputs.dashboard.jsonnet)
• Provides focused visibility into Job Inputs feature adoption and health
• Allows for future expansion as the feature matures

Option 2: Add to existing dashboard

• Could integrate into business-stats.dashboard.jsonnet (https://gitlab.com/gitlab-com/runbooks/-/blob/master/dashboards/ci-runners/business-stats.dashboard.jsonnet) or main.dashboard.jsonnet (https://gitlab.com/gitlab-com/runbooks/-/blob/master/dashboards/ci-runners/main.dashboard.jsonnet)
• Pros: Consolidates related metrics
• Cons: May dilute focus on this specific feature

Proposed Charts for New Dashboard

1. Interpolation Success Rate Over Time (Line chart)

   • Shows the ratio of successful interpolations vs. total attempts
   • Helps identify if error rates are increasing/decreasing
   • Breakdown by error type (parse, evaluation, sensitive value)

2. Interpolation Failure Rate Breakdown (Pie/Donut chart)

   • Shows distribution of failure types: parse errors, evaluation errors, unsupported sensitive values
   • Helps identify which error category needs attention
   • Can inform future improvements to error handling

Next Steps

• @timofurrer and @tmaczukin will meet next week to refine this
• Decide between Option 1 (new dashboard) vs Option 2 (add to existing)
• Refine chart selections based on observability needs
• Create the dashboard file in runbooks repository
• Add dashboard link to relevant documentation