refactor podAffinity generation (#2156)

Author: FxKu

charts/postgres-operator/crds/operatorconfigurations.yaml

@@ -278,12 +278,12 @@ spec:
                   pdb_name_format:
                     type: string
                     default: "postgres-{cluster}-pdb"
-                  pod_antiaffinity_topology_key:
-                    type: string
-                    default: "kubernetes.io/hostname"
                   pod_antiaffinity_preferred_during_scheduling:
                     type: boolean
                     default: false
+                  pod_antiaffinity_topology_key:
+                    type: string
+                    default: "kubernetes.io/hostname"
                   pod_environment_configmap:
                     type: string
                   pod_environment_secret:

charts/postgres-operator/values.yaml

@@ -165,10 +165,10 @@ configKubernetes:
 
   # defines the template for PDB (Pod Disruption Budget) names
   pdb_name_format: "postgres-{cluster}-pdb"
+  # switches pod anti affinity type to `preferredDuringSchedulingIgnoredDuringExecution`
+  pod_antiaffinity_preferred_during_scheduling: false
   # override topology key for pod anti affinity
   pod_antiaffinity_topology_key: "kubernetes.io/hostname"
-  # switches pod anti affinity type to `preferredDuringSchedulingIgnoredDuringExecution`
-  # pod_antiaffinity_preferred_during_scheduling: true
   # namespaced name of the ConfigMap with environment variables to populate on every pod
   # pod_environment_configmap: "default/my-custom-config"
   # name of the Secret (in cluster namespace) with environment variables to populate on every pod

manifests/configmap.yaml

@@ -109,6 +109,7 @@ data:
   # password_rotation_interval: "90"
   # password_rotation_user_retention: "180"
   pdb_name_format: "postgres-{cluster}-pdb"
+  # pod_antiaffinity_preferred_during_scheduling: "false"
   # pod_antiaffinity_topology_key: "kubernetes.io/hostname"
   pod_deletion_wait_timeout: 10m
   # pod_environment_configmap: "default/my-custom-config"

manifests/operatorconfiguration.crd.yaml

@@ -276,6 +276,9 @@ spec:
                   pdb_name_format:
                     type: string
                     default: "postgres-{cluster}-pdb"
+                  pod_antiaffinity_preferred_during_scheduling:
+                    type: boolean
+                    default: false
                   pod_antiaffinity_topology_key:
                     type: string
                     default: "kubernetes.io/hostname"

manifests/postgresql-operator-default-configuration.yaml

@@ -84,6 +84,7 @@ configuration:
     # node_readiness_label_merge: "OR"
     oauth_token_secret_name: postgresql-operator
     pdb_name_format: "postgres-{cluster}-pdb"
+    pod_antiaffinity_preferred_during_scheduling: false
     pod_antiaffinity_topology_key: "kubernetes.io/hostname"
     # pod_environment_configmap: "default/my-custom-config"
     # pod_environment_secret: "my-custom-secret"
@@ -95,6 +96,7 @@ configuration:
     # pod_service_account_role_binding_definition: ""
     pod_terminate_grace_period: 5m
     secret_name_template: "{username}.{cluster}.credentials.{tprkind}.{tprgroup}"
+    share_pgsocket_with_sidecars: false
     spilo_allow_privilege_escalation: true
     # spilo_runasuser: 101
     # spilo_runasgroup: 103

pkg/apis/acid.zalan.do/v1/crds.go

@@ -1372,12 +1372,12 @@ var OperatorConfigCRDResourceValidation = apiextv1.CustomResourceValidation{
 							"pdb_name_format": {
 								Type: "string",
 							},
-							"pod_antiaffinity_topology_key": {
-								Type: "string",
-							},
 							"pod_antiaffinity_preferred_during_scheduling": {
 								Type: "boolean",
 							},
+							"pod_antiaffinity_topology_key": {
+								Type: "string",
+							},
 							"pod_environment_configmap": {
 								Type: "string",
 							},

pkg/apis/acid.zalan.do/v1/operator_configuration_type.go

@@ -97,10 +97,10 @@ type KubernetesMetaConfiguration struct {
 	PodPriorityClassName                     string              `json:"pod_priority_class_name,omitempty"`
 	MasterPodMoveTimeout                     Duration            `json:"master_pod_move_timeout,omitempty"`
 	EnablePodAntiAffinity                    bool                `json:"enable_pod_antiaffinity,omitempty"`
-	PodAntiAffinityTopologyKey               string              `json:"pod_antiaffinity_topology_key,omitempty"`
 	PodAntiAffinityPreferredDuringScheduling bool                `json:"pod_antiaffinity_preferred_during_scheduling,omitempty"`
+	PodAntiAffinityTopologyKey               string              `json:"pod_antiaffinity_topology_key,omitempty"`
 	PodManagementPolicy                      string              `json:"pod_management_policy,omitempty"`
-  EnableReadinessProbe                     bool                `json:"enable_readiness_probe,omitempty"`
+	EnableReadinessProbe                     bool                `json:"enable_readiness_probe,omitempty"`
 	EnableCrossNamespaceSecret               bool                `json:"enable_cross_namespace_secret,omitempty"`
 }
 

pkg/cluster/connection_pooler.go

@@ -354,11 +354,12 @@ func (c *Cluster) generateConnectionPoolerPodTemplate(role PostgresRole) (
 	nodeAffinity := c.nodeAffinity(c.OpConfig.NodeReadinessLabel, spec.NodeAffinity)
 	if c.OpConfig.EnablePodAntiAffinity {
 		labelsSet := labels.Set(c.connectionPoolerLabels(role, false).MatchLabels)
-		podTemplate.Spec.Affinity = generatePodAffinity(
+		podTemplate.Spec.Affinity = podAffinity(
 			labelsSet,
 			c.OpConfig.PodAntiAffinityTopologyKey,
 			nodeAffinity,
 			c.OpConfig.PodAntiAffinityPreferredDuringScheduling,
+			true,
 		)
 	} else if nodeAffinity != nil {
 		podTemplate.Spec.Affinity = nodeAffinity

pkg/cluster/k8sres.go

@@ -495,8 +495,14 @@ func (c *Cluster) nodeAffinity(nodeReadinessLabel map[string]string, nodeAffinit
 	}
 }
 
-func generatePodAffinity(labels labels.Set, topologyKey string, nodeAffinity *v1.Affinity, preferredDuringScheduling bool) *v1.Affinity {
-	// generate pod anti-affinity to avoid multiple pods of the same Postgres cluster in the same topology , e.g. node
+func podAffinity(
+	labels labels.Set,
+	topologyKey string,
+	nodeAffinity *v1.Affinity,
+	preferredDuringScheduling bool,
+	anti bool) *v1.Affinity {
+
+	var podAffinity v1.Affinity
 
 	podAffinityTerm := v1.PodAffinityTerm{
 		LabelSelector: &metav1.LabelSelector{
@@ -505,24 +511,47 @@ func generatePodAffinity(labels labels.Set, topologyKey string, nodeAffinity *v1
 		TopologyKey: topologyKey,
 	}
 
-	podAffinity := v1.Affinity{
-		PodAntiAffinity: &v1.PodAntiAffinity{},
+	if anti {
+		podAffinity.PodAntiAffinity = generatePodAntiAffinity(podAffinityTerm, preferredDuringScheduling)
+	} else {
+		podAffinity.PodAffinity = generatePodAffinity(podAffinityTerm, preferredDuringScheduling)
 	}
 
+	if nodeAffinity != nil && nodeAffinity.NodeAffinity != nil {
+		podAffinity.NodeAffinity = nodeAffinity.NodeAffinity
+	}
+
+	return &podAffinity
+}
+
+func generatePodAffinity(podAffinityTerm v1.PodAffinityTerm, preferredDuringScheduling bool) *v1.PodAffinity {
+	podAffinity := &v1.PodAffinity{}
+
 	if preferredDuringScheduling {
-		podAffinity.PodAntiAffinity.PreferredDuringSchedulingIgnoredDuringExecution = []v1.WeightedPodAffinityTerm{{
+		podAffinity.PreferredDuringSchedulingIgnoredDuringExecution = []v1.WeightedPodAffinityTerm{{
 			Weight:          1,
 			PodAffinityTerm: podAffinityTerm,
 		}}
 	} else {
-		podAffinity.PodAntiAffinity.RequiredDuringSchedulingIgnoredDuringExecution = []v1.PodAffinityTerm{podAffinityTerm}
+		podAffinity.RequiredDuringSchedulingIgnoredDuringExecution = []v1.PodAffinityTerm{podAffinityTerm}
 	}
 
-	if nodeAffinity != nil && nodeAffinity.NodeAffinity != nil {
-		podAffinity.NodeAffinity = nodeAffinity.NodeAffinity
+	return podAffinity
+}
+
+func generatePodAntiAffinity(podAffinityTerm v1.PodAffinityTerm, preferredDuringScheduling bool) *v1.PodAntiAffinity {
+	podAntiAffinity := &v1.PodAntiAffinity{}
+
+	if preferredDuringScheduling {
+		podAntiAffinity.PreferredDuringSchedulingIgnoredDuringExecution = []v1.WeightedPodAffinityTerm{{
+			Weight:          1,
+			PodAffinityTerm: podAffinityTerm,
+		}}
+	} else {
+		podAntiAffinity.RequiredDuringSchedulingIgnoredDuringExecution = []v1.PodAffinityTerm{podAffinityTerm}
 	}
 
-	return &podAffinity
+	return podAntiAffinity
 }
 
 func tolerations(tolerationsSpec *[]v1.Toleration, podToleration map[string]string) []v1.Toleration {
@@ -778,11 +807,12 @@ func (c *Cluster) generatePodTemplate(
 	}
 
 	if podAntiAffinity {
-		podSpec.Affinity = generatePodAffinity(
+		podSpec.Affinity = podAffinity(
 			labels,
 			podAntiAffinityTopologyKey,
 			nodeAffinity,
 			podAntiAffinityPreferredDuringScheduling,
+			true,
 		)
 	} else if nodeAffinity != nil {
 		podSpec.Affinity = nodeAffinity
@@ -2100,20 +2130,15 @@ func (c *Cluster) generateLogicalBackupJob() (*batchv1.CronJob, error) {
 		c.OpConfig.ClusterNameLabel: c.Name,
 		"application":               "spilo-logical-backup",
 	}
-	podAffinityTerm := v1.PodAffinityTerm{
-		LabelSelector: &metav1.LabelSelector{
-			MatchLabels: labels,
-		},
-		TopologyKey: "kubernetes.io/hostname",
-	}
-	podAffinity := v1.Affinity{
-		PodAffinity: &v1.PodAffinity{
-			PreferredDuringSchedulingIgnoredDuringExecution: []v1.WeightedPodAffinityTerm{{
-				Weight:          1,
-				PodAffinityTerm: podAffinityTerm,
-			},
-			},
-		}}
+
+	nodeAffinity := c.nodeAffinity(c.OpConfig.NodeReadinessLabel, nil)
+	podAffinity := podAffinity(
+		labels,
+		"kubernetes.io/hostname",
+		nodeAffinity,
+		true,
+		false,
+	)
 
 	annotations := c.generatePodAnnotations(&c.Spec)
 
@@ -2147,7 +2172,7 @@ func (c *Cluster) generateLogicalBackupJob() (*batchv1.CronJob, error) {
 	}
 
 	// overwrite specific params of logical backups pods
-	podTemplate.Spec.Affinity = &podAffinity
+	podTemplate.Spec.Affinity = podAffinity
 	podTemplate.Spec.RestartPolicy = "Never" // affects containers within a pod
 
 	// configure a batch job