Got it, let's break this down. You already have a working ServiceMonitor to scrape the Blackbox Exporter's own metrics (from /metrics), but now you need to replicate that standalone Prometheus blackbox probe job using ServiceMonitor. Here's exactly how to do it:

Key Background #

The standalone blackbox job doesn't scrape the exporter directly—it tells the exporter to probe your targets by hitting its /probe endpoint with target parameters. So we need a separate ServiceMonitor for this probe task (since your existing one is for the exporter's internal metrics).

The ServiceMonitor Configuration #

Create this new ServiceMonitor (adjust values to match your cluster setup):

apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: blackbox-probe-servicemonitor
  labels:
    release: prometheus  # Match the release label of your kube-prometheus-stack installation
spec:
  # We don't need to select a Service here since we're using static targets
  selector:
    matchLabels:
      dummy: label  # A placeholder to avoid selector errors (can be any non-existent label)
  endpoints:
  - interval: 30s
    path: /probe  # Target the exporter's probe endpoint instead of /metrics
    port: http  # Match the port name from your Blackbox Exporter Service
    scrapeTimeout: 10s
    params:
      module: [http_2xx]  # Same module as your standalone config
    staticConfigs:
    - targets:
      - http://localhost:5000
      - http://localhost:3000
      - http://localhost:9090  # Your list of static targets to probe
    relabelings:  # Mirror the relabel rules from your standalone config
    - sourceLabels: [__address__]
      targetLabel: __param_target
    - sourceLabels: [__param_target]
      targetLabel: instance
    - targetLabel: __address__
      replacement: blackbox-exporter:9115  # Replace with your Blackbox Exporter Service address (e.g., <service-name>.<namespace>.svc.cluster.local:9115)

What Each Part Does #

Let's map this back to your original standalone Prometheus config:

• params: {module: [http_2xx]}: Exactly matches the params section in your old job.
• staticConfigs.targets: Your list of targets to probe, same as the static_configs in the standalone setup.
• relabelings: Directly translates your relabel_configs—these rules rewrite the target address to point to the Blackbox Exporter, pass the original target as a target parameter, and set the instance label to the target URL.
• release: prometheus: Ensures kube-prometheus-stack's Prometheus instance picks up this ServiceMonitor (adjust if your Helm release name is different, e.g., kube-prometheus-stack).

Critical Notes #

1. Service Address: The replacement value in the last relabel rule must be the cluster-internal address of your Blackbox Exporter Service. If the exporter is in the same namespace as this ServiceMonitor, just <service-name>:<port> works. Otherwise, use the full DNS name like blackbox-exporter.monitoring.svc.cluster.local:9115.
2. Port Name: The port: http must match the name field in your Blackbox Exporter Service's port definition (not just the port number).
3. Separate ServiceMonitors: Keep your original ServiceMonitor (for /metrics) and this new one (for /probe)—they serve different purposes: one tracks the exporter's health, the other runs your probes.

Once you apply this ServiceMonitor, check your Prometheus targets page—you should see a new job (named something like monitoring/blackbox-probe-servicemonitor/0) with your static targets being probed via the Blackbox Exporter.

内容的提问来源于stack exchange，提问作者Jonas