fix pod anti affinity (#144)

DataDog · Sep 2, 2020 · 4faeb97 · 4faeb97
1 parent be28a3d
commit 4faeb97
Show file tree

Hide file tree

Showing 5 changed files with 99 additions and 32 deletions.
diff --git a/pkg/controller/datadogagent/clusteragent.go b/pkg/controller/datadogagent/clusteragent.go
@@ -391,7 +391,7 @@ func newClusterAgentPodTemplate(agentdeployment *datadoghqv1alpha1.DatadogAgent,
 				VolumeMounts: volumeMounts,
 			},
 		},
-		Affinity:          getPodAffinity(clusterAgentSpec.Affinity, getClusterAgentName(agentdeployment)),
+		Affinity:          clusterAgentSpec.Affinity,
 		Tolerations:       clusterAgentSpec.Tolerations,
 		PriorityClassName: agentdeployment.Spec.ClusterAgent.PriorityClassName,
 		Volumes:           volumes,

diff --git a/pkg/controller/datadogagent/clusteragent_test.go b/pkg/controller/datadogagent/clusteragent_test.go
@@ -31,7 +31,6 @@ var testClusterAgentReplicas int32 = 1
 
 func clusterAgentDefaultPodSpec() corev1.PodSpec {
 	return corev1.PodSpec{
-		Affinity:           getPodAffinity(nil, "foo-cluster-agent"),
 		ServiceAccountName: "foo-cluster-agent",
 		Containers: []corev1.Container{
 			{
@@ -396,7 +395,7 @@ func Test_newClusterAgentDeploymentFromInstance_EnvVars(t *testing.T) {
 func Test_newClusterAgentDeploymentFromInstance_CustomDeploymentName(t *testing.T) {
 	customDeploymentName := "custom-cluster-agent-deployment"
 	deploymentNamePodSpec := clusterAgentDefaultPodSpec()
-	deploymentNamePodSpec.Affinity = getPodAffinity(nil, customDeploymentName)
+	deploymentNamePodSpec.Affinity = nil
 
 	deploymentNameAgentDeployment := test.NewDefaultedDatadogAgent("bar", "foo",
 		&test.NewDatadogAgentOptions{

diff --git a/pkg/controller/datadogagent/clusterchecksrunner.go b/pkg/controller/datadogagent/clusterchecksrunner.go
@@ -277,7 +277,7 @@ func newClusterChecksRunnerPodTemplate(dda *datadoghqv1alpha1.DatadogAgent, labe
 				},
 			},
 			Volumes:           getVolumesForClusterChecksRunner(dda),
-			Affinity:          getPodAffinity(clusterChecksRunnerSpec.Affinity, getClusterChecksRunnerName(dda)),
+			Affinity:          getPodAffinity(clusterChecksRunnerSpec.Affinity),
 			Tolerations:       clusterChecksRunnerSpec.Tolerations,
 			PriorityClassName: clusterChecksRunnerSpec.PriorityClassName,
 		},
@@ -479,3 +479,26 @@ func getVolumeMountsForClusterChecksRunner(dda *datadoghqv1alpha1.DatadogAgent)
 func getClusterChecksRunnerCustomConfigConfigMapName(dda *datadoghqv1alpha1.DatadogAgent) string {
 	return fmt.Sprintf("%s-runner-datadog-yaml", dda.Name)
 }
+
+// getPodAffinity returns the pod anti affinity of the cluster check runner pods
+// the default anti affinity ensures we don't schedule multiple cluster check runners on the same node
+func getPodAffinity(affinity *corev1.Affinity) *corev1.Affinity {
+	if affinity != nil {
+		return affinity
+	}
+
+	return &corev1.Affinity{
+		PodAntiAffinity: &corev1.PodAntiAffinity{
+			RequiredDuringSchedulingIgnoredDuringExecution: []corev1.PodAffinityTerm{
+				{
+					LabelSelector: &metav1.LabelSelector{
+						MatchLabels: map[string]string{
+							datadoghqv1alpha1.AgentDeploymentComponentLabelKey: datadoghqv1alpha1.DefaultClusterChecksRunnerResourceSuffix,
+						},
+					},
+					TopologyKey: "kubernetes.io/hostname",
+				},
+			},
+		},
+	}
+}
diff --git a/pkg/controller/datadogagent/clusterchecksrunner_test.go b/pkg/controller/datadogagent/clusterchecksrunner_test.go
@@ -2,26 +2,27 @@ package datadogagent
 
 import (
 	"fmt"
+	"reflect"
 	"testing"
 
-	appsv1 "k8s.io/api/apps/v1"
-	corev1 "k8s.io/api/core/v1"
-	apiequality "k8s.io/apimachinery/pkg/api/equality"
-	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	logf "sigs.k8s.io/controller-runtime/pkg/runtime/log"
-
 	datadoghqv1alpha1 "github.com/DataDog/datadog-operator/pkg/apis/datadoghq/v1alpha1"
 	test "github.com/DataDog/datadog-operator/pkg/apis/datadoghq/v1alpha1/test"
 	"github.com/DataDog/datadog-operator/pkg/controller/utils/comparison"
+
 	"github.com/google/go-cmp/cmp"
 	assert "github.com/stretchr/testify/require"
+	appsv1 "k8s.io/api/apps/v1"
+	corev1 "k8s.io/api/core/v1"
+	apiequality "k8s.io/apimachinery/pkg/api/equality"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	logf "sigs.k8s.io/controller-runtime/pkg/runtime/log"
 )
 
 var testClusterChecksRunnerReplicas int32 = 2
 
 func clusterChecksRunnerDefaultPodSpec() corev1.PodSpec {
 	return corev1.PodSpec{
-		Affinity:           getPodAffinity(nil, "foo-cluster-checks-runner"),
+		Affinity:           getPodAffinity(nil),
 		ServiceAccountName: "foo-cluster-checks-runner",
 		InitContainers: []corev1.Container{
 			{
@@ -399,3 +400,68 @@ func Test_newClusterChecksRunnerDeploymentFromInstance_EnvVars(t *testing.T) {
 	}
 	test.Run(t)
 }
+
+func Test_getPodAffinity(t *testing.T) {
+	tests := []struct {
+		name     string
+		affinity *corev1.Affinity
+		want     *corev1.Affinity
+	}{
+		{
+			name:     "no user-defined affinity - apply default",
+			affinity: nil,
+			want: &corev1.Affinity{
+				PodAntiAffinity: &corev1.PodAntiAffinity{
+					RequiredDuringSchedulingIgnoredDuringExecution: []corev1.PodAffinityTerm{
+						{
+							LabelSelector: &metav1.LabelSelector{
+								MatchLabels: map[string]string{
+									"agent.datadoghq.com/component": "cluster-checks-runner",
+								},
+							},
+							TopologyKey: "kubernetes.io/hostname",
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "user-defined affinity",
+			affinity: &corev1.Affinity{
+				PodAntiAffinity: &corev1.PodAntiAffinity{
+					RequiredDuringSchedulingIgnoredDuringExecution: []corev1.PodAffinityTerm{
+						{
+							LabelSelector: &metav1.LabelSelector{
+								MatchLabels: map[string]string{
+									"foo": "bar",
+								},
+							},
+							TopologyKey: "baz",
+						},
+					},
+				},
+			},
+			want: &corev1.Affinity{
+				PodAntiAffinity: &corev1.PodAntiAffinity{
+					RequiredDuringSchedulingIgnoredDuringExecution: []corev1.PodAffinityTerm{
+						{
+							LabelSelector: &metav1.LabelSelector{
+								MatchLabels: map[string]string{
+									"foo": "bar",
+								},
+							},
+							TopologyKey: "baz",
+						},
+					},
+				},
+			},
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := getPodAffinity(tt.affinity); !reflect.DeepEqual(got, tt.want) {
+				t.Errorf("getPodAffinity() = %v, want %v", got, tt.want)
+			}
+		})
+	}
+}
diff --git a/pkg/controller/datadogagent/utils.go b/pkg/controller/datadogagent/utils.go
@@ -1547,27 +1547,6 @@ func getDefaultAPMAgentLivenessProbe() *corev1.Probe {
 	return livenessProbe
 }
 
-func getPodAffinity(affinity *corev1.Affinity, labelValue string) *corev1.Affinity {
-	if affinity != nil {
-		return affinity
-	}
-
-	return &corev1.Affinity{
-		PodAntiAffinity: &corev1.PodAntiAffinity{
-			RequiredDuringSchedulingIgnoredDuringExecution: []corev1.PodAffinityTerm{
-				{
-					LabelSelector: &metav1.LabelSelector{
-						MatchLabels: map[string]string{
-							"app": labelValue,
-						},
-					},
-					TopologyKey: "kubernetes.io/hostname",
-				},
-			},
-		},
-	}
-}
-
 func updateDaemonSetStatus(ds *appsv1.DaemonSet, dsStatus *datadoghqv1alpha1.DaemonSetStatus, updateTime *metav1.Time) *datadoghqv1alpha1.DaemonSetStatus {
 	if dsStatus == nil {
 		dsStatus = &datadoghqv1alpha1.DaemonSetStatus{}