go-kratos/kratos · error

ErrorCode: %d

Error message

ErrorCode: %d

What it means

Thrown by Kratos' Bilibili Discovery registrar after the HTTP POST to the discovery server's /discovery/register endpoint succeeds at the transport layer, but the JSON response body carries a non-zero business code (res.Code != 0). The discovery server uses this code field to signal rejection of the registration (e.g. invalid appid/env, duplicate or conflicting instance registration). The wrapped value is only the numeric code; the server's message string is logged but not returned.

Source

Thrown at contrib/registry/discovery/impl_registrar.go:110

	} else {
		p.Set(_paramKeyStatus, strconv.FormatInt(ins.Status, 10))
	}
	p.Set(_paramKeyMetadata, string(metadata))

	// send request to Discovery server.
	if _, err = d.httpClient.R().
		SetContext(ctx).
		SetQueryParamsFromValues(p).
		SetResult(&res).
		Post(uri); err != nil {
		d.switchNode()
		log.Error("Discovery: register client.Get failed",
			"uri", uri+"?"+p.Encode(), "zone", c.Zone, "env", c.Env, "appid", ins.AppID, "addrs", ins.Addrs, "error", err)
		return
	}

	if res.Code != 0 {
		err = fmt.Errorf("ErrorCode: %d", res.Code)
		log.Error("Discovery: register client.Get returned code",
			"uri", uri, "env", c.Env, "appid", ins.AppID, "addrs", ins.Addrs, "code", res.Code)
	}

	log.Info("Discovery: register client.Get succeeded", "uri", uri, "env", c.Env, "appid", ins.AppID, "addrs", ins.Addrs)

	return
}

func (d *Discovery) Deregister(_ context.Context, service *registry.ServiceInstance) error {
	ins := fromServerInstance(service, d.config)
	return d.cancel(ins)
}

View on GitHub (pinned to 668db92c2c)

Solutions

1. Check the registrar log line 'Discovery: register client.Get returned code' for the uri/appid/code, and map that code against the Bilibili discovery server's code table (e.g. -304/-404 style codes) to see why registration was rejected
2. Verify the Discovery config (env, zone, appid, addrs) matches what is provisioned on the discovery server; appid naming rules and env values are strict
3. Confirm the addrs you register are reachable from the discovery server's health checks and correctly formatted (scheme://host:port)
4. If the server rejects due to stale state, deregister the old instance (or let its lease expire) and retry registration
5. If the code is unknown, curl the register URL from the logged uri+params directly and inspect the full {code,message} body for the exact reason

Example fix

// before: appid with invalid format gets rejected with ErrorCode: -400
cfg := discovery.WithAppID("My App")

// after: appid must match the server's naming rules
cfg := discovery.WithAppID("my.app.service")

// Pre-validate the instance shape before registering
func validInstance(ins *registry.ServiceInstance) error {
	if ins.Name == "" || len(ins.Endpoints) == 0 {
		return fmt.Errorf("invalid instance: name and endpoints required")
	}
	for _, ep := range ins.Endpoints {
		if u, err := url.Parse(ep); err != nil || u.Host == "" {
			return fmt.Errorf("invalid endpoint %q", ep)
		}
	}
	return nil
}

err := registrar.Register(ctx, ins)
if err != nil {
	if strings.HasPrefix(err.Error(), "ErrorCode:") {
		// business rejection from the discovery server: log code, do NOT blind-retry the same payload
		log.Error("discovery rejected registration", "err", err, "appid", ins.Name)
		return err
	}
	// transport error: the client already switched nodes; a later retry is safe
}

When it happens

Trigger: Calling registrar.Register()/registry.Register() with a Discovery instance: config env/zone values the server does not recognize, an AppID that is not provisioned on the discovery server, addrs that fail server-side validation, or re-registering an instance with conflicting metadata/status. The HTTP call itself must succeed (otherwise the resty error path at impl_registrar.go:98 is taken instead); only the parsed {code,message} body with code != 0 produces this error.

Common situations: Misconfigured KRATOS_DISCOVERY env/zone/appid in deployment config; pointing the registrar at a discovery cluster that does not know the appid; registering the same appid+addrs from a second environment causing a conflict code; discovery server version differences that return different rejection codes than the client expects.

Related errors

• RegisterInstance err %v,%v
• response Error %d
• retry after %d times
• unsupported key: %s format: %s
• invalid path: %q is not a message

AI-assisted analysis of go-kratos/kratos@668db92c2c (2026-08-16). Data as JSON: /api/errors/11a8f5378849b67a. Report an issue: GitHub.