health: remove self reporting of cluster/node name

Those metrics should be directly inferred by users' Prometheus config.
When Cilium installs ServiceMonitor we in fact already add nodes and we
can pretty much expect users with multiple clusters to add their own
label to differentiate clusters.

Signed-off-by: Arthur Outhenin-Chalandre <git@mrfreezeex.fr>

Author: MrFreezeex

Documentation/observability/metrics.rst

@@ -504,12 +504,12 @@ Name                                       Labels
 Node Connectivity
 ~~~~~~~~~~~~~~~~~
 
-============================================= ====================================================================================================================================================================== ========== ==================================================================================================================================================================================================================
-Name                                          Labels                                                                                                                                                                 Default    Description
-============================================= ====================================================================================================================================================================== ========== ==================================================================================================================================================================================================================
-``node_health_connectivity_status``           ``source_cluster``, ``source_node_name``, ``type``, ``status``                                                                                                         Enabled    Number of endpoints with last observed status of both ICMP and HTTP connectivity between the current Cilium agent and other Cilium nodes
-``node_health_connectivity_latency_seconds``  ``source_cluster``, ``source_node_name``, ``type``, ``address_type``, ``protocol``                                                                                     Enabled    Histogram of the last observed latency between the current Cilium agent and other Cilium nodes in seconds
-============================================= ====================================================================================================================================================================== ========== ==================================================================================================================================================================================================================
+============================================= ======================================== ========== ==========================================================================================================================================
+Name                                          Labels                                    Default    Description
+============================================= ======================================== ========== ==========================================================================================================================================
+``node_health_connectivity_status``           ``type``, ``status``                     Enabled    Number of endpoints with last observed status of both ICMP and HTTP connectivity between the current Cilium agent and other Cilium nodes
+``node_health_connectivity_latency_seconds``  ``type``, ``address_type``, ``protocol`` Enabled    Histogram of the last observed latency between the current Cilium agent and other Cilium nodes in seconds
+============================================= ======================================== ========== ==========================================================================================================================================
 
 Clustermesh
 ~~~~~~~~~~~

Documentation/operations/upgrade.rst

@@ -395,6 +395,11 @@ now report per cluster metric instead of a "global" count and were renamed to re
 * ``cilium_operator_clustermesh_remote_cluster_services``
 * ``cilium_operator_clustermesh_remote_cluster_service_exports``
 
+The following metrics no longer reports a ``source_cluster`` and a ``source_node_name`` label:
+* ``node_health_connectivity_status``
+* ``node_health_connectivity_latency_seconds``
+
+
 Deprecated Metrics
 ~~~~~~~~~~~~~~~~~~
 

pkg/health/server/server.go

@@ -6,7 +6,6 @@ package server
 import (
 	"fmt"
 	"log/slog"
-	"path"
 	"time"
 
 	"github.com/cilium/cilium/api/v1/client/daemon"
@@ -16,7 +15,6 @@ import (
 	"github.com/cilium/cilium/api/v1/models"
 	"github.com/cilium/cilium/pkg/api"
 	ciliumPkg "github.com/cilium/cilium/pkg/client"
-	ciliumDefaults "github.com/cilium/cilium/pkg/defaults"
 	healthClientPkg "github.com/cilium/cilium/pkg/health/client"
 	"github.com/cilium/cilium/pkg/health/defaults"
 	"github.com/cilium/cilium/pkg/health/probe"
@@ -171,8 +169,6 @@ func (s *Server) collectNodeConnectivityMetrics(report *healthReport) {
 	if s.localStatus == nil || report == nil {
 		return
 	}
-	localClusterName, localNodeName := getClusterNodeName(s.localStatus.Name)
-
 	endpointStatuses := make(map[healthClientPkg.ConnectivityStatusType]int)
 	nodeStatuses := make(map[healthClientPkg.ConnectivityStatusType]int)
 
@@ -214,70 +210,70 @@ func (s *Server) collectNodeConnectivityMetrics(report *healthReport) {
 		s.nodesSeen[n.Name] = struct{}{}
 
 		// HTTP endpoint primary
-		collectConnectivityMetric(s.logger, endpointPathStatus.PrimaryAddress.HTTP, localClusterName, localNodeName,
+		collectConnectivityMetric(s.logger, endpointPathStatus.PrimaryAddress.HTTP,
 			metrics.LabelPeerEndpoint, metrics.LabelTrafficHTTP, metrics.LabelAddressTypePrimary)
 
 		// HTTP endpoint secondary
 		for _, secondary := range endpointPathStatus.SecondaryAddresses {
-			collectConnectivityMetric(s.logger, secondary.HTTP, localClusterName, localNodeName,
+			collectConnectivityMetric(s.logger, secondary.HTTP,
 				metrics.LabelPeerEndpoint, metrics.LabelTrafficHTTP, metrics.LabelAddressTypeSecondary)
 		}
 
 		// HTTP node primary
-		collectConnectivityMetric(s.logger, nodePathPrimaryAddress.HTTP, localClusterName, localNodeName,
+		collectConnectivityMetric(s.logger, nodePathPrimaryAddress.HTTP,
 			metrics.LabelPeerNode, metrics.LabelTrafficHTTP, metrics.LabelAddressTypePrimary)
 
 		// HTTP node secondary
 		for _, secondary := range nodePathSecondaryAddress {
-			collectConnectivityMetric(s.logger, secondary.HTTP, localClusterName, localNodeName,
+			collectConnectivityMetric(s.logger, secondary.HTTP,
 				metrics.LabelPeerNode, metrics.LabelTrafficHTTP, metrics.LabelAddressTypeSecondary)
 		}
 
 		// ICMP endpoint primary
-		collectConnectivityMetric(s.logger, endpointPathStatus.PrimaryAddress.Icmp, localClusterName, localNodeName,
+		collectConnectivityMetric(s.logger, endpointPathStatus.PrimaryAddress.Icmp,
 			metrics.LabelPeerEndpoint, metrics.LabelTrafficICMP, metrics.LabelAddressTypePrimary)
 
 		// ICMP endpoint secondary
 		for _, secondary := range endpointPathStatus.SecondaryAddresses {
-			collectConnectivityMetric(s.logger, secondary.Icmp, localClusterName, localNodeName,
+			collectConnectivityMetric(s.logger, secondary.Icmp,
 				metrics.LabelPeerEndpoint, metrics.LabelTrafficICMP, metrics.LabelAddressTypeSecondary)
 		}
 
 		// ICMP node primary
-		collectConnectivityMetric(s.logger, nodePathPrimaryAddress.Icmp, localClusterName, localNodeName,
+		collectConnectivityMetric(s.logger, nodePathPrimaryAddress.Icmp,
 			metrics.LabelPeerNode, metrics.LabelTrafficICMP, metrics.LabelAddressTypePrimary)
 
 		// ICMP node secondary
 		for _, secondary := range nodePathSecondaryAddress {
-			collectConnectivityMetric(s.logger, secondary.Icmp, localClusterName, localNodeName,
+			collectConnectivityMetric(s.logger, secondary.Icmp,
 				metrics.LabelPeerNode, metrics.LabelTrafficICMP, metrics.LabelAddressTypeSecondary)
 		}
 	}
 
 	// Aggregated health statuses for endpoint connectivity
 	metrics.NodeHealthConnectivityStatus.WithLabelValues(
-		localClusterName, localNodeName, metrics.LabelPeerEndpoint, metrics.LabelReachable).
+		metrics.LabelPeerEndpoint, metrics.LabelReachable).
 		Set(float64(endpointStatuses[healthClientPkg.ConnStatusReachable]))
 
 	metrics.NodeHealthConnectivityStatus.WithLabelValues(
-		localClusterName, localNodeName, metrics.LabelPeerEndpoint, metrics.LabelUnreachable).
+		metrics.LabelPeerEndpoint, metrics.LabelUnreachable).
 		Set(float64(endpointStatuses[healthClientPkg.ConnStatusUnreachable]))
 
 	metrics.NodeHealthConnectivityStatus.WithLabelValues(
-		localClusterName, localNodeName, metrics.LabelPeerEndpoint, metrics.LabelUnknown).
+		metrics.LabelPeerEndpoint, metrics.LabelUnknown).
 		Set(float64(endpointStatuses[healthClientPkg.ConnStatusUnknown]))
 
 	// Aggregated health statuses for node connectivity
 	metrics.NodeHealthConnectivityStatus.WithLabelValues(
-		localClusterName, localNodeName, metrics.LabelPeerNode, metrics.LabelReachable).
+		metrics.LabelPeerNode, metrics.LabelReachable).
 		Set(float64(nodeStatuses[healthClientPkg.ConnStatusReachable]))
 
 	metrics.NodeHealthConnectivityStatus.WithLabelValues(
-		localClusterName, localNodeName, metrics.LabelPeerNode, metrics.LabelUnreachable).
+		metrics.LabelPeerNode, metrics.LabelUnreachable).
 		Set(float64(nodeStatuses[healthClientPkg.ConnStatusUnreachable]))
 
 	metrics.NodeHealthConnectivityStatus.WithLabelValues(
-		localClusterName, localNodeName, metrics.LabelPeerNode, metrics.LabelUnknown).
+		metrics.LabelPeerNode, metrics.LabelUnknown).
 		Set(float64(nodeStatuses[healthClientPkg.ConnStatusUnknown]))
 }
 
@@ -292,16 +288,6 @@ func collectConnectivityMetric(logger *slog.Logger, status *healthModels.Connect
 	}
 }
 
-// getClusterNodeName returns the cluster name and node name if possible.
-func getClusterNodeName(str string) (string, string) {
-	clusterName, nodeName := path.Split(str)
-	if len(clusterName) == 0 {
-		return ciliumDefaults.ClusterName, nodeName
-	}
-	// remove forward slash at the end if any for cluster name
-	return path.Dir(clusterName), nodeName
-}
-
 // GetStatusResponse returns the most recent cluster connectivity status.
 func (s *Server) GetStatusResponse() *healthModels.HealthStatusResponse {
 	s.RLock()