Merge branch 'main' into khewonc/introspection-docs-update

Makefile

@@ -91,7 +91,7 @@ $(KUSTOMIZE): Makefile  ## Download kustomize locally if necessary.
 
 ENVTEST = bin/$(PLATFORM)/setup-envtest
 $(ENVTEST): Makefile ## Download envtest-setup locally if necessary.
-	$(call go-get-tool,$@,sigs.k8s.io/controller-runtime/tools/setup-envtest@latest)
+	$(call go-get-tool,$@,sigs.k8s.io/controller-runtime/tools/setup-envtest@v0.0.0-20240320141353-395cfc7486e6)
 
 # go-get-tool will 'go get' any package $2 and install it to $1.
 PROJECT_DIR := $(shell dirname $(abspath $(lastword $(MAKEFILE_LIST))))

apis/datadoghq/v1alpha1/datadogmonitor_types.go

@@ -91,6 +91,8 @@ type DatadogMonitorOptions struct {
 	EvaluationDelay *int64 `json:"evaluationDelay,omitempty"`
 	// A Boolean indicating whether notifications from this monitor automatically inserts its triggering tags into the title.
 	IncludeTags *bool `json:"includeTags,omitempty"`
+	// A Boolean indicating whether the log alert monitor triggers a single alert or multiple alerts when any group breaches a threshold.
+	GroupbySimpleMonitor *bool `json:"groupbySimpleMonitor,omitempty"`
 	// Whether or not the monitor is locked (only editable by creator and admins).
 	Locked *bool `json:"locked,omitempty"`
 	// Time (in seconds) to allow a host to boot and applications to fully start before starting the evaluation of
@@ -104,6 +106,12 @@ type DatadogMonitorOptions struct {
 	NotificationPresetName DatadogMonitorOptionsNotificationPreset `json:"notificationPresetName,omitempty"`
 	// A Boolean indicating whether tagged users are notified on changes to this monitor.
 	NotifyAudit *bool `json:"notifyAudit,omitempty"`
+	// A string indicating the granularity a monitor alerts on. Only available for monitors with groupings.
+	// For instance, a monitor grouped by cluster, namespace, and pod can be configured to only notify on each new
+	// cluster violating the alert conditions by setting notify_by to ["cluster"]. Tags mentioned in notify_by must
+	// be a subset of the grouping tags in the query. For example, a query grouped by cluster and namespace cannot
+	// notify on region. Setting notify_by to [*] configures the monitor to notify as a simple-alert.
+	NotifyBy []string `json:"notifyBy,omitempty"`
 	// A Boolean indicating whether this monitor notifies when data stops reporting.
 	NotifyNoData *bool `json:"notifyNoData,omitempty"`
 	// An enum that controls how groups or monitors are treated if an evaluation does not return data points.
@@ -115,6 +123,8 @@ type DatadogMonitorOptions struct {
 	// The number of minutes after the last notification before a monitor re-notifies on the current status.
 	// It only re-notifies if it’s not resolved.
 	RenotifyInterval *int64 `json:"renotifyInterval,omitempty"`
+	// The number of times re-notification messages should be sent on the current status at the provided re-notification interval.
+	RenotifyOccurrences *int64 `json:"renotifyOccurrences,omitempty"`
 	// A Boolean indicating whether this monitor needs a full window of data before it’s evaluated. We highly
 	// recommend you set this to false for sparse metrics, otherwise some evaluations are skipped. Default is false.
 	RequireFullWindow *bool `json:"requireFullWindow,omitempty"`

config/crd/bases/v1/datadoghq.com_datadogmonitors.yaml

@@ -78,6 +78,9 @@ spec:
                       description: Time (in seconds) to delay evaluation, as a non-negative integer. For example, if the value is set to 300 (5min), the timeframe is set to last_5m and the time is 7:00, the monitor evaluates data from 6:50 to 6:55. This is useful for AWS CloudWatch and other backfilled metrics to ensure the monitor always has data during evaluation.
                       format: int64
                       type: integer
+                    groupbySimpleMonitor:
+                      description: A Boolean indicating whether the log alert monitor triggers a single alert or multiple alerts when any group breaches a threshold.
+                      type: boolean
                     includeTags:
                       description: A Boolean indicating whether notifications from this monitor automatically inserts its triggering tags into the title.
                       type: boolean
@@ -98,6 +101,11 @@ spec:
                     notifyAudit:
                       description: A Boolean indicating whether tagged users are notified on changes to this monitor.
                       type: boolean
+                    notifyBy:
+                      description: A string indicating the granularity a monitor alerts on. Only available for monitors with groupings. For instance, a monitor grouped by cluster, namespace, and pod can be configured to only notify on each new cluster violating the alert conditions by setting notify_by to ["cluster"]. Tags mentioned in notify_by must be a subset of the grouping tags in the query. For example, a query grouped by cluster and namespace cannot notify on region. Setting notify_by to [*] configures the monitor to notify as a simple-alert.
+                      items:
+                        type: string
+                      type: array
                     notifyNoData:
                       description: A Boolean indicating whether this monitor notifies when data stops reporting.
                       type: boolean
@@ -108,6 +116,10 @@ spec:
                       description: The number of minutes after the last notification before a monitor re-notifies on the current status. It only re-notifies if it’s not resolved.
                       format: int64
                       type: integer
+                    renotifyOccurrences:
+                      description: The number of times re-notification messages should be sent on the current status at the provided re-notification interval.
+                      format: int64
+                      type: integer
                     requireFullWindow:
                       description: A Boolean indicating whether this monitor needs a full window of data before it’s evaluated. We highly recommend you set this to false for sparse metrics, otherwise some evaluations are skipped. Default is false.
                       type: boolean

config/crd/bases/v1beta1/datadoghq.com_datadogmonitors.yaml

@@ -79,6 +79,9 @@ spec:
                   description: Time (in seconds) to delay evaluation, as a non-negative integer. For example, if the value is set to 300 (5min), the timeframe is set to last_5m and the time is 7:00, the monitor evaluates data from 6:50 to 6:55. This is useful for AWS CloudWatch and other backfilled metrics to ensure the monitor always has data during evaluation.
                   format: int64
                   type: integer
+                groupbySimpleMonitor:
+                  description: A Boolean indicating whether the log alert monitor triggers a single alert or multiple alerts when any group breaches a threshold.
+                  type: boolean
                 includeTags:
                   description: A Boolean indicating whether notifications from this monitor automatically inserts its triggering tags into the title.
                   type: boolean
@@ -99,6 +102,11 @@ spec:
                 notifyAudit:
                   description: A Boolean indicating whether tagged users are notified on changes to this monitor.
                   type: boolean
+                notifyBy:
+                  description: A string indicating the granularity a monitor alerts on. Only available for monitors with groupings. For instance, a monitor grouped by cluster, namespace, and pod can be configured to only notify on each new cluster violating the alert conditions by setting notify_by to ["cluster"]. Tags mentioned in notify_by must be a subset of the grouping tags in the query. For example, a query grouped by cluster and namespace cannot notify on region. Setting notify_by to [*] configures the monitor to notify as a simple-alert.
+                  items:
+                    type: string
+                  type: array
                 notifyNoData:
                   description: A Boolean indicating whether this monitor notifies when data stops reporting.
                   type: boolean
@@ -109,6 +117,10 @@ spec:
                   description: The number of minutes after the last notification before a monitor re-notifies on the current status. It only re-notifies if it’s not resolved.
                   format: int64
                   type: integer
+                renotifyOccurrences:
+                  description: The number of times re-notification messages should be sent on the current status at the provided re-notification interval.
+                  format: int64
+                  type: integer
                 requireFullWindow:
                   description: A Boolean indicating whether this monitor needs a full window of data before it’s evaluated. We highly recommend you set this to false for sparse metrics, otherwise some evaluations are skipped. Default is false.
                   type: boolean

controllers/datadogagent/feature/orchestratorexplorer/feature.go

@@ -16,11 +16,13 @@ import (
 	apiutils "github.com/DataDog/datadog-operator/apis/utils"
 	"github.com/DataDog/datadog-operator/pkg/controller/utils/comparison"
 	"github.com/DataDog/datadog-operator/pkg/kubernetes"
+	"github.com/DataDog/datadog-operator/pkg/utils"
 	"github.com/go-logr/logr"
 
 	apicommon "github.com/DataDog/datadog-operator/apis/datadoghq/common"
 	apicommonv1 "github.com/DataDog/datadog-operator/apis/datadoghq/common/v1"
 	common "github.com/DataDog/datadog-operator/controllers/datadogagent/common"
+	"github.com/DataDog/datadog-operator/controllers/datadogagent/component"
 	"github.com/DataDog/datadog-operator/controllers/datadogagent/feature"
 	"github.com/DataDog/datadog-operator/controllers/datadogagent/object"
 	"github.com/DataDog/datadog-operator/controllers/datadogagent/object/volume"
@@ -60,8 +62,12 @@ type orchestratorExplorerFeature struct {
 	logger                      logr.Logger
 	customConfigAnnotationKey   string
 	customConfigAnnotationValue string
+
+	processAgentRequired bool
 }
 
+const NoProcessAgentMinVersion = "7.51.0"
+
 // ID returns the ID of the Feature
 func (f *orchestratorExplorerFeature) ID() feature.IDType {
 	return feature.OrchestratorExplorerIDType
@@ -74,9 +80,19 @@ func (f *orchestratorExplorerFeature) Configure(dda *v2alpha1.DatadogAgent) (req
 
 	if orchestratorExplorer != nil && apiutils.BoolValue(orchestratorExplorer.Enabled) {
 		reqComp.ClusterAgent.IsRequired = apiutils.NewBoolPointer(true)
+		reqContainers := []apicommonv1.AgentContainerName{apicommonv1.CoreAgentContainerName}
+
+		// Process Agent is not required as of agent version 7.51.0
+		if nodeAgent, ok := dda.Spec.Override[v2alpha1.NodeAgentComponentName]; ok {
+			if nodeAgent.Image != nil && !utils.IsAboveMinVersion(component.GetAgentVersionFromImage(*nodeAgent.Image), NoProcessAgentMinVersion) {
+				f.processAgentRequired = true
+				reqContainers = append(reqContainers, apicommonv1.ProcessAgentContainerName)
+			}
+		}
+
 		reqComp.Agent = feature.RequiredComponent{
 			IsRequired: apiutils.NewBoolPointer(true),
-			Containers: []apicommonv1.AgentContainerName{apicommonv1.CoreAgentContainerName, apicommonv1.ProcessAgentContainerName},
+			Containers: reqContainers,
 		}
 
 		if orchestratorExplorer.Conf != nil {
@@ -116,6 +132,7 @@ func (f *orchestratorExplorerFeature) Configure(dda *v2alpha1.DatadogAgent) (req
 func (f *orchestratorExplorerFeature) ConfigureV1(dda *v1alpha1.DatadogAgent) (reqComp feature.RequiredComponents) {
 	f.owner = dda
 	orchestratorExplorer := dda.Spec.Features.OrchestratorExplorer
+	f.processAgentRequired = true
 
 	if orchestratorExplorer != nil && apiutils.BoolValue(orchestratorExplorer.Enabled) {
 		reqComp.ClusterAgent.IsRequired = apiutils.NewBoolPointer(true)
@@ -231,7 +248,9 @@ func (f *orchestratorExplorerFeature) ManageSingleContainerNodeAgent(managers fe
 // It should do nothing if the feature doesn't need to configure it.
 func (f *orchestratorExplorerFeature) ManageNodeAgent(managers feature.PodTemplateManagers, provider string) error {
 	for _, env := range f.getEnvVars() {
-		managers.EnvVar().AddEnvVarToContainer(apicommonv1.ProcessAgentContainerName, env)
+		if f.processAgentRequired {
+			managers.EnvVar().AddEnvVarToContainer(apicommonv1.ProcessAgentContainerName, env)
+		}
 		managers.EnvVar().AddEnvVarToContainer(apicommonv1.CoreAgentContainerName, env)
 	}
 

controllers/datadogagent/feature/orchestratorexplorer/feature_test.go

@@ -11,6 +11,7 @@ import (
 
 	apicommon "github.com/DataDog/datadog-operator/apis/datadoghq/common"
 	apicommonv1 "github.com/DataDog/datadog-operator/apis/datadoghq/common/v1"
+	"github.com/DataDog/datadog-operator/apis/datadoghq/v2alpha1"
 	v2alpha1test "github.com/DataDog/datadog-operator/apis/datadoghq/v2alpha1/test"
 
 	apiutils "github.com/DataDog/datadog-operator/apis/utils"
@@ -67,10 +68,11 @@ func Test_orchestratorExplorerFeature_Configure(t *testing.T) {
 				WithOrchestratorExplorerExtraTags([]string{"a:z", "b:y", "c:x"}).
 				WithOrchestratorExplorerDDUrl("https://foo.bar").
 				WithOrchestratorExplorerCustomConfigData(customConfDataV2).
+				WithComponentOverride(v2alpha1.NodeAgentComponentName, v2alpha1.DatadogAgentComponentOverride{Image: &apicommonv1.AgentImageConfig{Tag: "7.51.0"}}).
 				Build(),
 			WantConfigure: true,
 			ClusterAgent:  orchestratorExplorerClusterAgentWantFuncV2(),
-			Agent:         test.NewDefaultComponentTest().WithWantFunc(orchestratorExplorerNodeAgentWantFunc),
+			Agent:         test.NewDefaultComponentTest().WithWantFunc(orchestratorExplorerNodeAgentNoProcessAgentWantFunc),
 		},
 		{
 			Name: "v2alpha1 orchestrator explorer enabled and runs on cluster checks runner",
@@ -82,12 +84,27 @@ func Test_orchestratorExplorerFeature_Configure(t *testing.T) {
 				WithOrchestratorExplorerCustomConfigData(customConfDataV2).
 				WithClusterChecksEnabled(true).
 				WithClusterChecksUseCLCEnabled(true).
+				WithComponentOverride(v2alpha1.NodeAgentComponentName, v2alpha1.DatadogAgentComponentOverride{Image: &apicommonv1.AgentImageConfig{Tag: "7.51.0"}}).
 				Build(),
 			WantConfigure:       true,
 			ClusterAgent:        orchestratorExplorerClusterAgentWantFuncV2(),
-			Agent:               test.NewDefaultComponentTest().WithWantFunc(orchestratorExplorerNodeAgentWantFunc),
+			Agent:               test.NewDefaultComponentTest().WithWantFunc(orchestratorExplorerNodeAgentNoProcessAgentWantFunc),
 			ClusterChecksRunner: test.NewDefaultComponentTest().WithWantFunc(orchestratorExplorerClusterChecksRunnerWantFunc),
 		},
+		{
+			Name: "v2alpha1 orchestrator explorer enabled on version requiring process agent",
+			DDAv2: v2alpha1test.NewDatadogAgentBuilder().
+				WithOrchestratorExplorerEnabled(true).
+				WithOrchestratorExplorerScrubContainers(true).
+				WithOrchestratorExplorerExtraTags([]string{"a:z", "b:y", "c:x"}).
+				WithOrchestratorExplorerDDUrl("https://foo.bar").
+				WithOrchestratorExplorerCustomConfigData(customConfDataV2).
+				WithComponentOverride(v2alpha1.NodeAgentComponentName, v2alpha1.DatadogAgentComponentOverride{Image: &apicommonv1.AgentImageConfig{Tag: "7.50.0"}}).
+				Build(),
+			WantConfigure: true,
+			ClusterAgent:  orchestratorExplorerClusterAgentWantFuncV2(),
+			Agent:         test.NewDefaultComponentTest().WithWantFunc(orchestratorExplorerNodeAgentWantFunc),
+		},
 	}
 
 	tests.Run(t, buildOrchestratorExplorerFeature)
@@ -97,6 +114,16 @@ func orchestratorExplorerNodeAgentWantFunc(t testing.TB, mgrInterface feature.Po
 	mgr := mgrInterface.(*fake.PodTemplateManagers)
 	agentEnvVars := mgr.EnvVarMgr.EnvVarsByC[apicommonv1.ProcessAgentContainerName]
 	assert.True(t, apiutils.IsEqualStruct(agentEnvVars, expectedOrchestratorEnvsV2), "Process agent envvars \ndiff = %s", cmp.Diff(agentEnvVars, expectedOrchestratorEnvsV2))
+	agentEnvVars = mgr.EnvVarMgr.EnvVarsByC[apicommonv1.CoreAgentContainerName]
+	assert.True(t, apiutils.IsEqualStruct(agentEnvVars, expectedOrchestratorEnvsV2), "Core agent envvars \ndiff = %s", cmp.Diff(agentEnvVars, expectedOrchestratorEnvsV2))
+}
+
+func orchestratorExplorerNodeAgentNoProcessAgentWantFunc(t testing.TB, mgrInterface feature.PodTemplateManagers) {
+	mgr := mgrInterface.(*fake.PodTemplateManagers)
+	agentEnvVars := mgr.EnvVarMgr.EnvVarsByC[apicommonv1.ProcessAgentContainerName]
+	assert.True(t, apiutils.IsEqualStruct(agentEnvVars, nil), "Process agent envvars \ndiff = %s", cmp.Diff(agentEnvVars, expectedOrchestratorEnvsV2))
+	agentEnvVars = mgr.EnvVarMgr.EnvVarsByC[apicommonv1.CoreAgentContainerName]
+	assert.True(t, apiutils.IsEqualStruct(agentEnvVars, expectedOrchestratorEnvsV2), "Core agent envvars \ndiff = %s", cmp.Diff(agentEnvVars, expectedOrchestratorEnvsV2))
 }
 
 func orchestratorExplorerClusterChecksRunnerWantFunc(t testing.TB, mgrInterface feature.PodTemplateManagers) {