Add mode label to failed_jobs_total and job_duration_seconds metrics

Summary

• Add a mode label (steps/traditional) to gitlab_runner_failed_jobs_total and gitlab_runner_job_duration_seconds metrics
• Enables the Steps rollout monitoring dashboard to directly compare failure rates and job durations between execution modes
• Without this, failure/duration attribution requires instance-based proxy comparisons (comparing enabled vs non-enabled managers)

What changed

File	Change
common/build.go	Export DispatchedJobExecutionMode(), pass mode in JobFailureData on failure
common/trace.go	Add Mode field to JobFailureData
common/network.go	Update FailuresCollector interface to accept mode
helpers/prometheus/failures_collector.go	Add mode to metric labels, permutation, and recording
network/trace.go	Pass mode through to RecordFailure
commands/builds_helper.go	Add mode label to jobDurationHistogram
common/mocks.go	Update mock to match new interface

New metric labels

gitlab_runner_failed_jobs_total: added mode (values: steps, traditional, or empty for pre-mode-determination failures)

gitlab_runner_job_duration_seconds: added mode (values: steps, traditional)

Related

• Steps rollout dashboard: gitlab-com/runbooks!10340
• Ring 1 deployment: gitlab-com/gl-infra/chef-repo!7162
• Rollout plan: step-runner#319

Test plan

• TestFailuresCollector_Collect_GroupingReasons passes
• TestFailuresCollector_Collect_MetricsValues passes — verifies mode label is emitted
• go build ./... passes
• CI pipeline

Closes Add execution mode label to failure and duratio... (step-runner#445 - closed) • Georgi N. Georgiev | GitLab • 19.0