Merge pull request #1742 from flavio-fernandes/merge-06-jul-23

SDN-3993: [DownstreamMerge] 06 jul 23
openshift · Jul 7, 2023 · d917250 · d917250
2 parents 1458500 + 2f8e12e
commit d917250
Show file tree

Hide file tree

Showing 28 changed files with 975 additions and 251 deletions.
diff --git a/dist/templates/ovnkube-alerts.yaml.j2 b/dist/templates/ovnkube-alerts.yaml.j2
@@ -19,15 +19,15 @@ spec:
        description: There is no running ovnkube-manager
 
     - alert: OvnKubeManagerMultipleLeaders
-      expr: sum(ovnkube_master_leader) > 1
+      expr: sum(ovnkube_controller_leader) > 1
       for: 4m
       labels:
        severity: critical
       annotations:
        description: There are multiple ovnkube-manager leaders
 
     - alert: OvnKubeManagerNoLeader
-      expr: max(ovnkube_master_leader) == 0
+      expr: max(ovnkube_controller_leader) == 0
       for: 4m
       labels:
         severity: critical
@@ -61,7 +61,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
          sum by (le) (
-           rate(ovnkube_master_pod_creation_latency_seconds_bucket[15m])
+           rate(ovnkube_controller_pod_creation_latency_seconds_bucket[15m])
          ) 
         ) > 5
       for: 10m
@@ -76,14 +76,14 @@ spec:
       expr: |
         histogram_quantile(0.99,
          sum by (le) ( 
-          rate(ovnkube_master_pod_creation_latency_seconds_bucket[15m] 
+          rate(ovnkube_controller_pod_creation_latency_seconds_bucket[15m] 
           offset 15m)
          )
         )
         -
         histogram_quantile(0.99,
          sum by (le) (
-          rate(ovnkube_master_pod_creation_latency_seconds_bucket[15m])
+          rate(ovnkube_controller_pod_creation_latency_seconds_bucket[15m])
          )
         ) > 5 
 
@@ -99,7 +99,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
           sum by (le) (
-            rate(ovnkube_master_ovn_cli_latency_seconds_bucket{
+            rate(ovnkube_controller_ovn_cli_latency_seconds_bucket{
               command="ovn-nbctl"}[15m])
           )
         ) > 3 
@@ -115,7 +115,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
           sum by (le) (
-            rate(ovnkube_master_ovn_cli_latency_seconds_bucket{
+            rate(ovnkube_controller_ovn_cli_latency_seconds_bucket{
               command="ovn-sbctl"}[15m])
           )
         ) > 3
@@ -131,7 +131,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
           sum by (le) (
-            rate(ovnkube_master_resource_update_latency_seconds_bucket{
+            rate(ovnkube_controller_resource_update_latency_seconds_bucket{
               name="NetworkPolicy"}[15m])
           )
         ) > 1
@@ -147,7 +147,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
           sum by (le) (
-            rate(ovnkube_master_resource_update_latency_seconds_bucket{
+            rate(ovnkube_controller_resource_update_latency_seconds_bucket{
               name="Namespace"}[15m])
           )
         ) > 1
@@ -163,7 +163,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
           sum by (le) (
-            rate(ovnkube_master_resource_update_latency_seconds_bucket{
+            rate(ovnkube_controller_resource_update_latency_seconds_bucket{
               name="Service"}[15m])
           )
         ) > 1
@@ -179,7 +179,7 @@ spec:
       expr: |
         histogram_quantile(0.99,
           sum by (le) (
-            rate(ovnkube_master_resource_update_latency_seconds_bucket{
+            rate(ovnkube_controller_resource_update_latency_seconds_bucket{
               name="Endpoint"}[15m])
           )
         ) > 1
@@ -192,7 +192,7 @@ spec:
          across all masters for the last 15minutes is more than 1 seconds.
 
     - alert: OvnNBDBStale
-      expr: time() - max(ovnkube_master_nb_e2e_timestamp) > 120
+      expr: time() - max(ovnkube_controller_nb_e2e_timestamp) > 120
       for: 4m
       labels:
        severity: critical
@@ -201,7 +201,7 @@ spec:
          ovn-kubernetes has not written anything to the northbound database for too long
 
     - alert: OvnSBDBStale
-      expr: max(ovnkube_master_nb_e2e_timestamp) - max(ovnkube_master_sb_e2e_timestamp) > 120
+      expr: max(ovnkube_controller_nb_e2e_timestamp) - max(ovnkube_controller_sb_e2e_timestamp) > 120
       for: 4m
       labels:
        severity: critical

diff --git a/docs/metrics.md b/docs/metrics.md
@@ -18,6 +18,7 @@ Measurement accuracy can be impacted by other parallel processing that might be
 This list is to help notify if there are additions, changes or removals to metrics. Latest changes are at the top of this list.
 
 - Effect of OVN IC architecture:
+  - Move all the metrics from subsystem "ovnkube-master" to subsystem "ovnkube-controller". The non-IC and IC deployments will each continue to have their ovnkube-master and ovnkube-controller containers running inside the ovnkube-master and ovnkube-controller pods. The metrics scraping should work seemlessly. See https://github.com/ovn-org/ovn-kubernetes/pull/3723 for details
   - Move the following metrics from subsystem "master" to subsystem "clustermanager". Therefore, the follow metrics are renamed.
     - `ovnkube_master_num_v4_host_subnets` -> `ovnkube_clustermanager_num_v4_host_subnets`
     - `ovnkube_master_num_v6_host_subnets` -> `ovnkube_clustermanager_num_v6_host_subnets`

diff --git a/go-controller/cmd/ovnkube/ovnkube.go b/go-controller/cmd/ovnkube/ovnkube.go
@@ -301,9 +301,9 @@ func startOvnKube(ctx *cli.Context, cancel context.CancelFunc) error {
 		return runOvnKube(ctx.Context, runMode, ovnClientset, eventRecorder)
 	}
 
-	// Register prometheus metrics that do not depend on becoming ovnkube-master
-	// leader and get the proper HA config depending on the mode. For network
-	// manager mode or combined cluster and network manager modes (the classic
+	// Register prometheus metrics that do not depend on becoming ovnkube-controller
+	// leader and get the proper HA config depending on the mode. For ovnkube
+	// controller mode or combined cluster manager and ovnkube-controller modes (the classic
 	// master mode), the master HA config applies. For cluster manager
 	// standalone mode, the cluster manager HA config applies.
 	var haConfig *config.HAConfig
@@ -313,7 +313,7 @@ func startOvnKube(ctx *cli.Context, cancel context.CancelFunc) error {
 		metrics.RegisterClusterManagerBase()
 		fallthrough
 	case runMode.ovnkubeController:
-		metrics.RegisterMasterBase()
+		metrics.RegisterOVNKubeControllerBase()
 		haConfig = &config.MasterHA
 		name = networkControllerManagerLockName()
 	case runMode.clusterManager:
@@ -483,7 +483,7 @@ func runOvnKube(ctx context.Context, runMode *ovnkubeRunMode, ovnClientset *util
 		defer cm.Stop()
 
 		// record delay until ready
-		metrics.MetricMasterReadyDuration.Set(time.Since(startTime).Seconds())
+		metrics.MetricOVNKubeControllerReadyDuration.Set(time.Since(startTime).Seconds())
 	}
 
 	if runMode.node {
@@ -544,22 +544,22 @@ func runOvnKube(ctx context.Context, runMode *ovnkubeRunMode, ovnClientset *util
 	return nil
 }
 
-type ovnkubeMasterMetrics struct {
+type leaderMetrics struct {
 	runMode *ovnkubeRunMode
 }
 
-func (m ovnkubeMasterMetrics) On(string) {
+func (m leaderMetrics) On(string) {
 	if m.runMode.ovnkubeController {
-		metrics.MetricMasterLeader.Set(1)
+		metrics.MetricOVNKubeControllerLeader.Set(1)
 	}
 	if m.runMode.clusterManager {
 		metrics.MetricClusterManagerLeader.Set(1)
 	}
 }
 
-func (m ovnkubeMasterMetrics) Off(string) {
+func (m leaderMetrics) Off(string) {
 	if m.runMode.ovnkubeController {
-		metrics.MetricMasterLeader.Set(0)
+		metrics.MetricOVNKubeControllerLeader.Set(0)
 	}
 	if m.runMode.clusterManager {
 		metrics.MetricClusterManagerLeader.Set(0)
@@ -571,7 +571,7 @@ type ovnkubeMetricsProvider struct {
 }
 
 func (p ovnkubeMetricsProvider) NewLeaderMetric() leaderelection.SwitchMetric {
-	return &ovnkubeMasterMetrics{p.runMode}
+	return &leaderMetrics{p.runMode}
 }
 
 func networkControllerManagerLockName() string {

diff --git a/...lustermanager/clustermanager_suit_test.go → ...ustermanager/clustermanager_suite_test.go b/...lustermanager/clustermanager_suit_test.go → ...ustermanager/clustermanager_suite_test.go
diff --git a/go-controller/pkg/metrics/cluster_manager.go b/go-controller/pkg/metrics/cluster_manager.go
@@ -11,7 +11,7 @@ import (
 
 var registerClusterManagerBaseMetrics sync.Once
 
-// MetricMasterLeader identifies whether this instance of ovnkube-master is a leader or not
+// MetricClusterManagerLeader identifies whether this instance of ovnkube-cluster-manager is a leader or not
 var MetricClusterManagerLeader = prometheus.NewGauge(prometheus.GaugeOpts{
 	Namespace: MetricOvnkubeNamespace,
 	Subsystem: MetricOvnkubeSubsystemClusterManager,

diff --git a/go-controller/pkg/metrics/metrics.go b/go-controller/pkg/metrics/metrics.go
@@ -26,7 +26,7 @@ import (
 
 const (
 	MetricOvnkubeNamespace               = "ovnkube"
-	MetricOvnkubeSubsystemMaster         = "master"
+	MetricOvnkubeSubsystemController     = "controller"
 	MetricOvnkubeSubsystemClusterManager = "clustermanager"
 	MetricOvnkubeSubsystemNode           = "node"
 	MetricOvnNamespace                   = "ovn"