Advanced Metrics

Deep dive into monitoring Orchestrix workflows with professional metrics tools.

Key Metrics to Track

For any production workflow, you should track at least these four Golden Signals:

1. Latency: Total flow duration and individual step durations.
2. Traffic: Number of flow executions per second/minute.
3. Errors: Failure rates for flows and steps.
4. Saturation: If running many parallel steps, monitor your event loop or thread pool.

Implementation with Prometheus

import { Counter, Histogram } from 'prom-client';

const flowLatency = new Histogram({
  name: 'orchestrix_flow_duration_seconds',
  help: 'Duration of flows in seconds',
  labelNames: ['flow_name', 'status']
});

const stepRetries = new Counter({
  name: 'orchestrix_step_retries_total',
  help: 'Total number of step retries',
  labelNames: ['flow_name', 'step_name']
});

const flow = create("my-flow", {
  hooks: {
    onFlowComplete: ({ result }) => {
      flowLatency.observe({ 
        flow_name: result.name, 
        status: result.status 
      }, result.durationMs / 1000);
    },
    onStepRetry: (event, attempt) => {
      stepRetries.inc({ 
        flow_name: event.flowName, 
        step_name: event.stepName 
      });
    }
  }
});

Dashboard Ideas

Workflow Health

• Success Rate %: sum(rate(orchestrix_flow_duration_seconds_count{status="completed"}[5m])) / sum(rate(orchestrix_flow_duration_seconds_count[5m]))
• P99 Latency: histogram_quantile(0.99, sum by (le) (rate(orchestrix_flow_duration_seconds_bucket[5m])))

Step Reliability

• Top 5 Retrying Steps: Identify flaky dependencies.
• Compensation Rate: How often do your flows fail and require undoing?

Business Metrics

Don't forget that workflows often represent business processes. You can use the FlowContext in your hooks to extract business-level metrics:

onFlowSuccess: (result) => {
  const amount = result.context.get<number>('totalAmount');
  revenueCounter.inc(amount);
}