Fractal Techware

Alert runbooks /

BlackboxProbeFailed

A synthetic check run by blackbox_exporter against an endpoint has been failing continuously.

   
Severity critical
Source blackbox_exporter /probe (http, tcp, dns, icmp modules)
Key metric probe_success (1 = check passed, 0 = failed)

What it means

Prometheus asks blackbox_exporter to probe a target with a module, and the exporter reports probe_success. The alert fires when a probe has stayed at 0 for several minutes, so a single timeout does not page you.

A failed probe is what your users see from the exporter’s vantage point: the site is down, the port is closed, the certificate is invalid or the response did not match what the module expects. It can also be a problem with the probe itself, so verify before escalating.

Common causes

First checks

  1. List failing targets:
    probe_success == 0
    
  2. Look at what part failed:
    probe_http_status_code{instance="<target>"}
    

    and probe_http_duration_seconds by phase (resolve, connect, tls, processing, transfer).

  3. Run the probe with full debug output; it shows DNS, connection, TLS and response details plus the reason it failed:
    curl -s "http://<blackbox-exporter>:9115/probe?target=<target>&module=<module>&debug=true"
    
  4. Reproduce from outside the exporter:
    curl -sv -o /dev/null https://<target>/
    
  5. If only the exporter fails, check preferred_ip_protocol in the module and network policies around the exporter pod.

Fixing it

If the service is down, this is an incident for the owning team; the probe is doing its job. If the service changed (new redirect, new status code, new body), update the module to match. If only the exporter’s network path is broken, fix egress or run the exporter closer to the target.