feat(kuma-cp) health check add event log support (#1631)

kumahq · Mar 5, 2021 · d6685c3 · d6685c3
1 parent 3a380ca
commit d6685c3
Show file tree

Hide file tree

Showing 8 changed files with 240 additions and 79 deletions.
diff --git a/api/mesh/v1alpha1/health_check.pb.go b/api/mesh/v1alpha1/health_check.pb.go
diff --git a/api/mesh/v1alpha1/health_check.proto b/api/mesh/v1alpha1/health_check.proto
@@ -66,6 +66,25 @@ message HealthCheck {
     // service.
     google.protobuf.BoolValue fail_traffic_on_panic = 11;
 
+    // Specifies the path to the file where Envoy can log health check events.
+    // If empty, no event log will be written.
+    string event_log_path = 12;
+
+    // If set to true, health check failure events will always be logged. If set
+    // to false, only the initial health check failure event will be logged. The
+    // default value is false.
+    google.protobuf.BoolValue always_log_health_check_failures = 13;
+
+    // The "no traffic interval" is a special health check interval that is used
+    // when a cluster has never had traffic routed to it. This lower interval
+    // allows cluster information to be kept up to date, without sending a
+    // potentially large amount of active health checking traffic for no reason.
+    // Once a cluster has been used for traffic routing, Envoy will shift back
+    // to using the standard health check interval that is defined. Note that
+    // this interval takes precedence over any other. The default value for "no
+    // traffic interval" is 60 seconds.
+    google.protobuf.Duration no_traffic_interval = 14;
+
     // Tcp defines optional configuration for specifying bytes to send and
     // expected response during the health check
     message Tcp {

diff --git a/pkg/core/resources/apis/mesh/healthcheck_validator.go b/pkg/core/resources/apis/mesh/healthcheck_validator.go
@@ -105,6 +105,9 @@ func (d *HealthCheckResource) validateConf() (err validators.ValidationError) {
 	if d.Spec.Conf.IntervalJitter != nil {
 		err.Add(ValidateDuration(path.Field("intervalJitter"), d.Spec.Conf.IntervalJitter))
 	}
+	if d.Spec.Conf.NoTrafficInterval != nil {
+		err.Add(ValidateDuration(path.Field("noTrafficInterval"), d.Spec.Conf.NoTrafficInterval))
+	}
 	err.Add(d.validatePercentage(path.Field("healthyPanicThreshold"), d.Spec.Conf.HealthyPanicThreshold))
 	if d.Spec.Conf.GetHttp() != nil {
 		err.Add(d.validateConfHttp(path.Field("http")))

diff --git a/pkg/core/resources/apis/mesh/healthcheck_validator_test.go b/pkg/core/resources/apis/mesh/healthcheck_validator_test.go
@@ -114,6 +114,7 @@ var _ = Describe("HealthCheck", func() {
                   initialJitter: 0s
                   intervalJitter: 0s
                   healthyPanicThreshold: 101
+                  noTrafficInterval: 0s
                   http:
                     path: ""
                     requestHeadersToAdd:
@@ -138,6 +139,8 @@ var _ = Describe("HealthCheck", func() {
                   message: must have a positive value
                 - field: conf.intervalJitter
                   message: must have a positive value
+                - field: conf.noTrafficInterval
+                  message: must have a positive value
                 - field: conf.healthyPanicThreshold
                   message: must be in range [0.0 - 100.0]
                 - field: conf.http.path

diff --git a/pkg/xds/envoy/clusters/v2/health_check_configurer.go b/pkg/xds/envoy/clusters/v2/health_check_configurer.go
@@ -139,13 +139,16 @@ func (e *HealthCheckConfigurer) Configure(cluster *envoy_api.Cluster) error {
 		HealthChecker: &envoy_core.HealthCheck_TcpHealthCheck_{
 			TcpHealthCheck: &envoy_core.HealthCheck_TcpHealthCheck{},
 		},
-		Interval:              activeChecks.Interval,
-		Timeout:               activeChecks.Timeout,
-		UnhealthyThreshold:    &wrappers.UInt32Value{Value: activeChecks.UnhealthyThreshold},
-		HealthyThreshold:      &wrappers.UInt32Value{Value: activeChecks.HealthyThreshold},
-		InitialJitter:         activeChecks.InitialJitter,
-		IntervalJitter:        activeChecks.IntervalJitter,
-		IntervalJitterPercent: activeChecks.IntervalJitterPercent,
+		Interval:                     activeChecks.Interval,
+		Timeout:                      activeChecks.Timeout,
+		UnhealthyThreshold:           &wrappers.UInt32Value{Value: activeChecks.UnhealthyThreshold},
+		HealthyThreshold:             &wrappers.UInt32Value{Value: activeChecks.HealthyThreshold},
+		InitialJitter:                activeChecks.InitialJitter,
+		IntervalJitter:               activeChecks.IntervalJitter,
+		IntervalJitterPercent:        activeChecks.IntervalJitterPercent,
+		EventLogPath:                 activeChecks.EventLogPath,
+		AlwaysLogHealthCheckFailures: activeChecks.AlwaysLogHealthCheckFailures.GetValue(),
+		NoTrafficInterval:            activeChecks.NoTrafficInterval,
 	}
 
 	healthPanicThreshold(cluster, activeChecks.GetHealthyPanicThreshold())

diff --git a/pkg/xds/envoy/clusters/v2/health_check_configurer_test.go b/pkg/xds/envoy/clusters/v2/health_check_configurer_test.go
@@ -348,6 +348,44 @@ var _ = Describe("HealthCheckConfigurer", func() {
               timeout: 4s
               unhealthyThreshold: 3
             name: testCluster
+            type: EDS`,
+		}),
+		Entry("HealthCheck with event log path", testCase{
+			clusterName: "testCluster",
+			healthCheck: &mesh_core.HealthCheckResource{
+				Spec: &mesh_proto.HealthCheck{
+					Sources: []*mesh_proto.Selector{
+						{Match: mesh_proto.TagSelector{"kuma.io/service": "backend"}},
+					},
+					Destinations: []*mesh_proto.Selector{
+						{Match: mesh_proto.TagSelector{"kuma.io/service": "redis"}},
+					},
+					Conf: &mesh_proto.HealthCheck_Conf{
+						Interval:                     ptypes.DurationProto(5 * time.Second),
+						Timeout:                      ptypes.DurationProto(4 * time.Second),
+						NoTrafficInterval:            ptypes.DurationProto(6 * time.Second),
+						UnhealthyThreshold:           3,
+						HealthyThreshold:             2,
+						EventLogPath:                 "/event/log/path",
+						AlwaysLogHealthCheckFailures: &wrappers.BoolValue{Value: true},
+					},
+				},
+			},
+			expected: `
+            connectTimeout: 5s
+            edsClusterConfig:
+              edsConfig:
+                ads: {}
+            healthChecks:
+            - alwaysLogHealthCheckFailures: true
+              eventLogPath: /event/log/path
+              healthyThreshold: 2
+              interval: 5s
+              noTrafficInterval: 6s
+              tcpHealthCheck: {}
+              timeout: 4s
+              unhealthyThreshold: 3
+            name: testCluster
             type: EDS`,
 		}),
 	)

diff --git a/pkg/xds/envoy/clusters/v3/health_check_configurer.go b/pkg/xds/envoy/clusters/v3/health_check_configurer.go
@@ -141,13 +141,16 @@ func (e *HealthCheckConfigurer) Configure(cluster *envoy_cluster.Cluster) error
 		HealthChecker: &envoy_core.HealthCheck_TcpHealthCheck_{
 			TcpHealthCheck: &envoy_core.HealthCheck_TcpHealthCheck{},
 		},
-		Interval:              activeChecks.Interval,
-		Timeout:               activeChecks.Timeout,
-		UnhealthyThreshold:    &wrappers.UInt32Value{Value: activeChecks.UnhealthyThreshold},
-		HealthyThreshold:      &wrappers.UInt32Value{Value: activeChecks.HealthyThreshold},
-		InitialJitter:         activeChecks.InitialJitter,
-		IntervalJitter:        activeChecks.IntervalJitter,
-		IntervalJitterPercent: activeChecks.IntervalJitterPercent,
+		Interval:                     activeChecks.Interval,
+		Timeout:                      activeChecks.Timeout,
+		UnhealthyThreshold:           &wrappers.UInt32Value{Value: activeChecks.UnhealthyThreshold},
+		HealthyThreshold:             &wrappers.UInt32Value{Value: activeChecks.HealthyThreshold},
+		InitialJitter:                activeChecks.InitialJitter,
+		IntervalJitter:               activeChecks.IntervalJitter,
+		IntervalJitterPercent:        activeChecks.IntervalJitterPercent,
+		EventLogPath:                 activeChecks.EventLogPath,
+		AlwaysLogHealthCheckFailures: activeChecks.AlwaysLogHealthCheckFailures.GetValue(),
+		NoTrafficInterval:            activeChecks.NoTrafficInterval,
 	}
 
 	healthPanicThreshold(cluster, activeChecks.GetHealthyPanicThreshold())

diff --git a/pkg/xds/envoy/clusters/v3/health_check_configurer_test.go b/pkg/xds/envoy/clusters/v3/health_check_configurer_test.go
@@ -356,6 +356,45 @@ var _ = Describe("HealthCheckConfigurer", func() {
               timeout: 4s
               unhealthyThreshold: 3
             name: testCluster
+            type: EDS`,
+		}),
+		Entry("HealthCheck with event log path", testCase{
+			clusterName: "testCluster",
+			healthCheck: &mesh_core.HealthCheckResource{
+				Spec: &mesh_proto.HealthCheck{
+					Sources: []*mesh_proto.Selector{
+						{Match: mesh_proto.TagSelector{"kuma.io/service": "backend"}},
+					},
+					Destinations: []*mesh_proto.Selector{
+						{Match: mesh_proto.TagSelector{"kuma.io/service": "redis"}},
+					},
+					Conf: &mesh_proto.HealthCheck_Conf{
+						Interval:                     ptypes.DurationProto(5 * time.Second),
+						Timeout:                      ptypes.DurationProto(4 * time.Second),
+						NoTrafficInterval:            ptypes.DurationProto(6 * time.Second),
+						UnhealthyThreshold:           3,
+						HealthyThreshold:             2,
+						EventLogPath:                 "/event/log/path",
+						AlwaysLogHealthCheckFailures: &wrappers.BoolValue{Value: true},
+					},
+				},
+			},
+			expected: `
+            connectTimeout: 5s
+            edsClusterConfig:
+              edsConfig:
+                ads: {}
+                resourceApiVersion: V3
+            healthChecks:
+            - alwaysLogHealthCheckFailures: true
+              eventLogPath: /event/log/path
+              healthyThreshold: 2
+              interval: 5s
+              noTrafficInterval: 6s
+              tcpHealthCheck: {}
+              timeout: 4s
+              unhealthyThreshold: 3
+            name: testCluster
             type: EDS`,
 		}),
 	)