tsenart/vegeta · error

error registering prometheus metrics: %s

Error message

error registering prometheus metrics: %s

What it means

When -prometheus-addr is set, attack() creates a prometheus registry and registers the metrics collector; if pm.Register(r) fails, the error is wrapped. Registration fails when the same metric name or duplicate collectors are registered, or malformed metric descriptors are used.

Source

Thrown at attack.go:194

	out, err := file(opts.outputf, true)
	if err != nil {
		return fmt.Errorf("error opening %s: %s", opts.outputf, err)
	}
	defer out.Close()

	tlsc, err := tlsConfig(opts.insecure, opts.certf, opts.keyf, opts.rootCerts)
	if err != nil {
		return err
	}

	var pm *prom.Metrics
	if opts.promAddr != "" {
		pm = prom.NewMetrics()

		r := prometheus.NewRegistry()
		if err := pm.Register(r); err != nil {
			return fmt.Errorf("error registering prometheus metrics: %s", err)
		}

		srv := http.Server{
			Addr:    opts.promAddr,
			Handler: prom.NewHandler(r, time.Now().UTC()),
		}

		defer srv.Close()
		go srv.ListenAndServe()
	}

	atk := vegeta.NewAttacker(
		vegeta.Redirects(opts.redirects),
		vegeta.Timeout(opts.timeout),
		vegeta.LocalAddr(*opts.laddr.IPAddr),
		vegeta.TLSConfig(tlsc),
		vegeta.Workers(opts.workers),
		vegeta.MaxWorkers(opts.maxWorkers),

View on GitHub (pinned to cf58112690)

Solutions

1. Use a fresh prometheus.NewRegistry() per attack run (as upstream does) rather than a shared global registry.
2. Check the wrapped error for 'already registered' and remove the duplicate collector registration.
3. Verify the prometheus client_golang version matches what the metric definitions expect (metric/label name rules changed across versions).

reg := prometheus.NewRegistry()
if err := prom.NewMetrics().Register(reg); err != nil {
    return fmt.Errorf("prometheus registration pre-check failed: %w", err)
}

if err := run(); err != nil {
    if strings.Contains(err.Error(), "already registered") {
        log.Fatalf("duplicate metric registration: use a fresh registry per run: %v", err)
    }
    log.Fatal(err)
}

When it happens

Trigger: `-prometheus-addr` supplied and prom.Metrics.Register returns an error from prometheus.Register — typically duplicate metric registration or an invalid metric name/label configuration in the vendored prom version.

Common situations: Custom builds where Metrics is instantiated twice against one registry; dependency upgrades changing prometheus client metric naming rules; embedding attack() into custom code that shares a registry.

Related errors

• failed to register metric %v: %w
• rate frequency and time unit must be bigger than zero
• bad certificate
• -rate=0 requires setting -max-workers
• format %q isn't one of [%s]

AI-assisted analysis of tsenart/vegeta@cf58112690 (2026-08-31). Data as JSON: /api/errors/924cfab6f83526df. Report an issue: GitHub.