BCF-2641 Add prom bridge_name metric label to PromPipelineTasksTotalF…

…inished (#10861) * Add prom bridge_name metric label to PromPipelineTasksTotalFinished * Update changelog * Update changelog * Add prom pipeline_tasks_total_fin... metric label addition to changelog
smartcontractkit · Oct 23, 2023 · ae8ec42 · ae8ec42
1 parent 5211322
commit ae8ec42
Show file tree

Hide file tree

Showing 3 changed files with 10 additions and 3 deletions.
diff --git a/core/services/fluxmonitorv2/flux_monitor.go b/core/services/fluxmonitorv2/flux_monitor.go
@@ -1040,7 +1040,7 @@ func (fm *FluxMonitor) isValidSubmission(l logger.Logger, answer decimal.Decimal
 	pipeline.PromPipelineTaskExecutionTime.WithLabelValues(fmt.Sprintf("%d", jobId), jobName, "", job.FluxMonitor.String()).Set(float64(elapsed))
 	pipeline.PromPipelineRunErrors.WithLabelValues(fmt.Sprintf("%d", jobId), jobName).Inc()
 	pipeline.PromPipelineRunTotalTimeToCompletion.WithLabelValues(fmt.Sprintf("%d", jobId), jobName).Set(float64(elapsed))
-	pipeline.PromPipelineTasksTotalFinished.WithLabelValues(fmt.Sprintf("%d", jobId), jobName, "", job.FluxMonitor.String(), "error").Inc()
+	pipeline.PromPipelineTasksTotalFinished.WithLabelValues(fmt.Sprintf("%d", jobId), jobName, "", job.FluxMonitor.String(), "", "error").Inc()
 	return false
 }
 

diff --git a/core/services/pipeline/runner.go b/core/services/pipeline/runner.go
@@ -98,7 +98,7 @@ var (
 		Name: "pipeline_tasks_total_finished",
 		Help: "The total number of pipeline tasks which have finished",
 	},
-		[]string{"job_id", "job_name", "task_id", "task_type", "status"},
+		[]string{"job_id", "job_name", "task_id", "task_type", "bridge_name", "status"},
 	)
 )
 
@@ -488,7 +488,13 @@ func logTaskRunToPrometheus(trr TaskRunResult, spec Spec) {
 	} else {
 		status = "completed"
 	}
-	PromPipelineTasksTotalFinished.WithLabelValues(fmt.Sprintf("%d", spec.JobID), spec.JobName, trr.Task.DotID(), string(trr.Task.Type()), status).Inc()
+
+	bridgeName := ""
+	if bridgeTask, ok := trr.Task.(*BridgeTask); ok {
+		bridgeName = bridgeTask.Name
+	}
+
+	PromPipelineTasksTotalFinished.WithLabelValues(fmt.Sprintf("%d", spec.JobID), spec.JobName, trr.Task.DotID(), string(trr.Task.Type()), bridgeName, status).Inc()
 }
 
 // ExecuteAndInsertFinishedRun executes a run in memory then inserts the finished run/task run records, returning the final result

diff --git a/docs/CHANGELOG.md b/docs/CHANGELOG.md
@@ -21,6 +21,7 @@ URL = '...'
 ServerPubKey = '...'
 ```
 These will eventually replace `TelemetryIngress.URL` and `TelemetryIngress.ServerPubKey`. Setting `TelemetryIngress.URL` and `TelemetryIngress.ServerPubKey` alongside `[[TelemetryIngress.Endpoints]]` will prevent the node from booting. Only one way of configuring telemetry endpoints is supported.
+- Added bridge_name label to `pipeline_tasks_total_finished` prometheus metric. This should make it easier to see directly what bridge was failing out from the CL NODE perspective.
 
 ### Upcoming Required Configuration Change