如何配置Istio Sidecar控制上报至Mixer的数据？含健康URL排除

Got it, this is a super common pain point—health checks fire constantly and cluttering your metrics with them doesn’t add any meaningful insight. Let’s walk through how to configure Istio to skip reporting these requests to Mixer (or the newer telemetry pipeline, since Istio 1.5+ uses a lighter-weight setup).

Using Istio Telemetry v2 (Recommended for Istio 1.5+) #

The modern way to control telemetry in Istio is with the Telemetry custom resource. You can create a rule that excludes your /health endpoint from metrics collection entirely.

Here’s a sample configuration you can apply:

apiVersion: telemetry.istio.io/v1alpha1
kind: Telemetry
metadata:
  name: skip-health-check-metrics
  namespace: istio-system
spec:
  metrics:
  - providers:
    - name: prometheus
    overrides:
    # Disable metrics specifically for /health requests
    - match:
        metric: REQUEST_DURATION
        match: "request.url_path == '/health'"
      disabled: true
    # Add more overrides if you need to exclude other health paths like /healthz or /ready
    - match:
        metric: REQUEST_DURATION
        match: "request.url_path == '/healthz' OR request.url_path == '/ready'"
      disabled: true
  selector:
    matchLabels:
      istio: ingressgateway # Adjust this to target specific workloads (e.g., app: your-service)

Let me break down what this does:

• We’re targeting the REQUEST_DURATION metric (the main request latency metric Istio collects)
• The match condition identifies requests to your health check paths
• Setting disabled: true tells the sidecar not to report these requests to Prometheus/Mixer

Applying the Config #

Save this as skip-health-metrics.yaml and run:

kubectl apply -f skip-health-metrics.yaml

Istio will automatically push this config to all sidecars matching the selector. Give it 30-60 seconds for the changes to take effect.

For Older Istio Versions (1.4 and Below) #

If you’re still using the legacy Mixer setup, you’ll need to use Mixer rules to filter out health check requests:

1. Create a rule that routes health check requests to a "noop" handler (which discards them):

apiVersion: config.istio.io/v1alpha2
kind: rule
metadata:
  name: skip-health-check
  namespace: istio-system
spec:
  match: "request.url_path == '/health'"
  actions:
  - handler: noop
    instances: []

2. Apply this rule, and ensure your existing metrics rules don’t override this filter. The noop handler will absorb the health check requests without sending them to your metrics system.

Verifying the Setup #

After applying the config:

• Check your Prometheus metrics (e.g., istio_request_duration_seconds_bucket) and confirm there are no entries where request_url_path is /health for your service.
• You can also run istioctl proxy-config listeners <your-pod-name> to verify the sidecar has picked up the updated telemetry rules.

内容的提问来源于stack exchange，提问作者David