Configure Apache Airflow monitoring with Prometheus alerts and Grafana dashboards

Install monitoring dependencies

Install the required packages for Prometheus, StatsD, and monitoring tools.

sudo apt update
sudo apt install -y prometheus prometheus-node-exporter statsd python3-statsd
sudo apt install -y grafana

sudo dnf install -y epel-release
sudo dnf install -y prometheus node_exporter statsd python3-statsd
sudo dnf install -y grafana

Install Airflow monitoring dependencies

Install the Python packages needed for Airflow metrics export and StatsD integration.

pip install apache-airflow[statsd,prometheus]
pip install prometheus_client statsd

Configure Airflow for StatsD metrics

Update the Airflow configuration to enable StatsD metrics export for monitoring DAG and task performance.

[metrics]
statsd_on = True
statsd_host = localhost
statsd_port = 8125
statsd_prefix = airflow
statsd_allow_list = 
statsd_custom_client_path = 

[scheduler]
statsd_on = True
statsd_host = localhost
statsd_port = 8125
statsd_prefix = airflow.scheduler

[webserver]
statsd_on = True
statsd_host = localhost
statsd_port = 8125
statsd_prefix = airflow.webserver

Create StatsD configuration

Configure StatsD to collect Airflow metrics and export them in Prometheus format.

{
  graphitePort: 2003
, graphiteHost: "127.0.0.1"
, port: 8125
, backends: [ "./backends/prometheus" ]
, prometheus: {
    prefix: "airflow_",
    port: 9102
  }
, deleteIdleStats: true
, deleteGauges: true
, deleteTimers: true
, deleteSets: true
, deleteCounters: true
}

Configure Airflow metrics endpoint

Create a custom metrics endpoint for Airflow to expose Prometheus-compatible metrics directly.

from airflow.plugins_manager import AirflowPlugin
from airflow.www.app import csrf
from flask import Blueprint, Response
from prometheus_client import generate_latest, CONTENT_TYPE_LATEST
from airflow.models import DagRun, TaskInstance, DagModel
from airflow.utils.db import provide_session
from prometheus_client import Gauge, Counter
import logging

log = logging.getLogger(__name__)

Define metrics
dag_run_total = Counter('airflow_dag_runs_total', 'Total DAG runs', ['dag_id', 'state'])
task_instance_total = Counter('airflow_task_instances_total', 'Total task instances', ['dag_id', 'task_id', 'state'])
dag_run_duration = Gauge('airflow_dag_run_duration_seconds', 'DAG run duration', ['dag_id'])
task_duration = Gauge('airflow_task_duration_seconds', 'Task duration', ['dag_id', 'task_id'])

prometheus_blueprint = Blueprint('prometheus', __name__)

@prometheus_blueprint.route('/metrics')
@provide_session
def metrics(session=None):
    """Prometheus metrics endpoint"""
    try:
        # Update DAG run metrics
        dag_runs = session.query(DagRun).all()
        for dag_run in dag_runs:
            dag_run_total.labels(dag_id=dag_run.dag_id, state=dag_run.state)._value._value += 1
            if dag_run.end_date and dag_run.start_date:
                duration = (dag_run.end_date - dag_run.start_date).total_seconds()
                dag_run_duration.labels(dag_id=dag_run.dag_id).set(duration)
        
        # Update task instance metrics
        task_instances = session.query(TaskInstance).all()
        for ti in task_instances:
            task_instance_total.labels(dag_id=ti.dag_id, task_id=ti.task_id, state=ti.state)._value._value += 1
            if ti.end_date and ti.start_date:
                duration = (ti.end_date - ti.start_date).total_seconds()
                task_duration.labels(dag_id=ti.dag_id, task_id=ti.task_id).set(duration)
        
        return Response(generate_latest(), mimetype=CONTENT_TYPE_LATEST)
    except Exception as e:
        log.error(f"Error generating metrics: {e}")
        return Response("Error generating metrics", status=500)

class PrometheusMetricsPlugin(AirflowPlugin):
    name = "prometheus_metrics"
    flask_blueprints = [prometheus_blueprint]

Configure Prometheus targets

Add Airflow metrics endpoints to Prometheus scraping configuration for data collection.

global:
  scrape_interval: 15s
  evaluation_interval: 15s

rule_files:
  - "airflow_alerts.yml"

alerting:
  alertmanagers:
    - static_configs:
        - targets:
          - localhost:9093

scrape_configs:
  - job_name: 'prometheus'
    static_configs:
      - targets: ['localhost:9090']

  - job_name: 'node-exporter'
    static_configs:
      - targets: ['localhost:9100']

  - job_name: 'airflow-webserver'
    static_configs:
      - targets: ['localhost:8080']
    metrics_path: '/metrics'
    scrape_interval: 30s

  - job_name: 'airflow-statsd'
    static_configs:
      - targets: ['localhost:9102']
    scrape_interval: 15s

  - job_name: 'airflow-scheduler'
    static_configs:
      - targets: ['localhost:8793']
    metrics_path: '/health'
    scrape_interval: 30s

Create Airflow alerting rules

Define comprehensive alerting rules for DAG failures, task timeouts, and system health issues.

groups:
name: airflow_alerts
  rules:
  - alert: AirflowDagFailed
    expr: increase(airflow_dag_runs_total{state="failed"}[5m]) > 0
    for: 1m
    labels:
      severity: critical
    annotations:
      summary: "Airflow DAG {{ $labels.dag_id }} failed"
      description: "DAG {{ $labels.dag_id }} has failed runs in the last 5 minutes"

  - alert: AirflowTaskFailed
    expr: increase(airflow_task_instances_total{state="failed"}[5m]) > 0
    for: 2m
    labels:
      severity: warning
    annotations:
      summary: "Airflow task {{ $labels.task_id }} in DAG {{ $labels.dag_id }} failed"
      description: "Task {{ $labels.task_id }} in DAG {{ $labels.dag_id }} has failed"

  - alert: AirflowSchedulerDown
    expr: up{job="airflow-scheduler"} == 0
    for: 1m
    labels:
      severity: critical
    annotations:
      summary: "Airflow scheduler is down"
      description: "Airflow scheduler has been down for more than 1 minute"

  - alert: AirflowWebserverDown
    expr: up{job="airflow-webserver"} == 0
    for: 2m
    labels:
      severity: warning
    annotations:
      summary: "Airflow webserver is down"
      description: "Airflow webserver has been down for more than 2 minutes"

  - alert: AirflowDagRunDuration
    expr: airflow_dag_run_duration_seconds > 3600
    for: 5m
    labels:
      severity: warning
    annotations:
      summary: "Airflow DAG {{ $labels.dag_id }} running too long"
      description: "DAG {{ $labels.dag_id }} has been running for more than 1 hour"

  - alert: AirflowTaskQueueHigh
    expr: airflow_executor_queued_tasks > 100
    for: 5m
    labels:
      severity: warning
    annotations:
      summary: "High number of queued Airflow tasks"
      description: "More than 100 tasks are queued in Airflow executor"

  - alert: AirflowDagImportErrors
    expr: airflow_dag_processing_import_errors > 0
    for: 1m
    labels:
      severity: critical
    annotations:
      summary: "Airflow DAG import errors detected"
      description: "{{ $value }} DAG import errors detected in Airflow"

Start monitoring services

Enable and start Prometheus, StatsD, and Grafana services for monitoring collection.

sudo systemctl enable --now prometheus
sudo systemctl enable --now node_exporter
sudo systemctl enable --now statsd
sudo systemctl enable --now grafana-server

Restart Airflow services with new configuration
sudo systemctl restart airflow-webserver
sudo systemctl restart airflow-scheduler

Configure Grafana data source

Add Prometheus as a data source in Grafana for Airflow metrics visualization.

apiVersion: 1

datasources:
  - name: Prometheus
    type: prometheus
    access: proxy
    url: http://localhost:9090
    isDefault: true
    editable: false

Create Airflow Grafana dashboard

Import a comprehensive Airflow dashboard with DAG metrics, task performance, and system health panels.

{
  "dashboard": {
    "id": null,
    "title": "Apache Airflow Monitoring",
    "tags": ["airflow"],
    "timezone": "browser",
    "panels": [
      {
        "id": 1,
        "title": "DAG Runs by State",
        "type": "stat",
        "targets": [
          {
            "expr": "sum by (state) (airflow_dag_runs_total)",
            "legendFormat": "{{state}}"
          }
        ],
        "gridPos": {"h": 8, "w": 12, "x": 0, "y": 0}
      },
      {
        "id": 2,
        "title": "Task Instances by State",
        "type": "piechart",
        "targets": [
          {
            "expr": "sum by (state) (airflow_task_instances_total)",
            "legendFormat": "{{state}}"
          }
        ],
        "gridPos": {"h": 8, "w": 12, "x": 12, "y": 0}
      },
      {
        "id": 3,
        "title": "DAG Run Duration",
        "type": "graph",
        "targets": [
          {
            "expr": "airflow_dag_run_duration_seconds",
            "legendFormat": "{{dag_id}}"
          }
        ],
        "gridPos": {"h": 8, "w": 24, "x": 0, "y": 8}
      },
      {
        "id": 4,
        "title": "Failed DAGs (Last 24h)",
        "type": "table",
        "targets": [
          {
            "expr": "increase(airflow_dag_runs_total{state=\"failed\"}[24h]) > 0",
            "format": "table"
          }
        ],
        "gridPos": {"h": 8, "w": 12, "x": 0, "y": 16}
      },
      {
        "id": 5,
        "title": "System Resources",
        "type": "graph",
        "targets": [
          {
            "expr": "100 - (avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m])) * 100)",
            "legendFormat": "CPU Usage %"
          },
          {
            "expr": "(1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)) * 100",
            "legendFormat": "Memory Usage %"
          }
        ],
        "gridPos": {"h": 8, "w": 12, "x": 12, "y": 16}
      }
    ],
    "time": {"from": "now-24h", "to": "now"},
    "refresh": "30s"
  }
}

Configure Grafana dashboard provisioning

Set up automatic dashboard provisioning to load the Airflow monitoring dashboard on startup.

apiVersion: 1

providers:
  - name: 'airflow'
    orgId: 1
    folder: 'Airflow'
    type: file
    disableDeletion: false
    editable: true
    updateIntervalSeconds: 10
    allowUiUpdates: true
    options:
      path: /etc/grafana/provisioning/dashboards

Set up Grafana alerts

Configure Grafana notification channels and alert rules for Airflow monitoring.

notifiers:
  - name: email-alerts
    type: email
    uid: email-alerts
    orgId: 1
    isDefault: true
    settings:
      addresses: "admin@example.com"
      subject: "Airflow Alert"
      uploadImage: true

Restart services with new configuration

Restart all services to apply the monitoring configuration and begin collecting metrics.

sudo systemctl restart prometheus
sudo systemctl restart grafana-server
sudo systemctl restart statsd

Verify services are running
sudo systemctl status prometheus grafana-server statsd