Merge pull request #13278 from Automaat/feat/extend-kuma-sd-api

kuma_sd: Extend Kuma SD configuration to allow users to specify ClientId

config/config_test.go

@@ -568,6 +568,7 @@ var expectedConf = &Config{
 			ServiceDiscoveryConfigs: discovery.Configs{
 				&xds.KumaSDConfig{
 					Server:           "http://kuma-control-plane.kuma-system.svc:5676",
+					ClientID:         "main-prometheus",
 					HTTPClientConfig: config.DefaultHTTPClientConfig,
 					RefreshInterval:  model.Duration(15 * time.Second),
 					FetchTimeout:     model.Duration(2 * time.Minute),

config/testdata/conf.good.yml

@@ -221,6 +221,7 @@ scrape_configs:
 
     kuma_sd_configs:
       - server: http://kuma-control-plane.kuma-system.svc:5676
+        client_id: main-prometheus
 
   - job_name: service-marathon
     marathon_sd_configs:

config/testdata/roundtrip.good.yml

@@ -108,6 +108,7 @@ scrape_configs:
 
     kuma_sd_configs:
       - server: http://kuma-control-plane.kuma-system.svc:5676
+        client_id: main-prometheus
 
     marathon_sd_configs:
       - servers:

discovery/xds/kuma.go

@@ -155,10 +155,14 @@ func kumaMadsV1ResourceParser(resources []*anypb.Any, typeURL string) ([]model.L
 
 func NewKumaHTTPDiscovery(conf *KumaSDConfig, logger log.Logger, reg prometheus.Registerer) (discovery.Discoverer, error) {
 	// Default to "prometheus" if hostname is unavailable.
-	clientID, err := osutil.GetFQDN()
-	if err != nil {
-		level.Debug(logger).Log("msg", "error getting FQDN", "err", err)
-		clientID = "prometheus"
+	clientID := conf.ClientID
+	if clientID == "" {
+		var err error
+		clientID, err = osutil.GetFQDN()
+		if err != nil {
+			level.Debug(logger).Log("msg", "error getting FQDN", "err", err)
+			clientID = "prometheus"
+		}
 	}
 
 	clientConfig := &HTTPResourceClientConfig{

discovery/xds/kuma_test.go

@@ -205,7 +205,7 @@ func TestNewKumaHTTPDiscovery(t *testing.T) {
 	require.True(t, ok)
 	require.Equal(t, kumaConf.Server, resClient.Server())
 	require.Equal(t, KumaMadsV1ResourceTypeURL, resClient.ResourceTypeURL())
-	require.NotEmpty(t, resClient.ID())
+	require.Equal(t, kumaConf.ClientID, resClient.ID())
 	require.Equal(t, KumaMadsV1ResourceType, resClient.config.ResourceType)
 }
 

discovery/xds/xds.go

@@ -55,6 +55,7 @@ type SDConfig struct {
 	RefreshInterval  model.Duration          `yaml:"refresh_interval,omitempty"`
 	FetchTimeout     model.Duration          `yaml:"fetch_timeout,omitempty"`
 	Server           string                  `yaml:"server,omitempty"`
+	ClientID         string                  `yaml:"client_id,omitempty"`
 }
 
 // mustRegisterMessage registers the provided message type in the typeRegistry, and panics

discovery/xds/xds_test.go

@@ -36,6 +36,7 @@ var (
 	sdConf = SDConfig{
 		Server:          "http://127.0.0.1",
 		RefreshInterval: model.Duration(10 * time.Second),
+		ClientID:        "test-id",
 	}
 
 	testFetchFailuresCount = prometheus.NewCounter(

docs/configuration/configuration.md

@@ -2230,6 +2230,11 @@ See below for the configuration options for Kuma MonitoringAssignment discovery:
 # Address of the Kuma Control Plane's MADS xDS server.
 server: <string>
 
+# Client id is used by Kuma Control Plane to compute Monitoring Assignment for specific Prometheus backend. 
+# This is useful when migrating between multiple Prometheus backends, or having separate backend for each Mesh.
+# When not specified, system hostname/fqdn will be used if available, if not `prometheus` will be used.
+[ client_id: <string> ]
+
 # The time to wait between polling update requests.
 [ refresh_interval: <duration> | default = 30s ]