Set statefulset update and management policy explicitly (zalando#515)

sdudoladov · web-flow · commit 0b53dbe5dcea · 2019-03-13T11:49:18.000+01:00
* fix logging in retry

* explicitly set the stateful set update strategy to onDelete

* add podManagementPolicy
diff --git a/docs/reference/operator_parameters.md b/docs/reference/operator_parameters.md
@@ -226,6 +226,11 @@ configuration they are grouped under the `kubernetes` key.
   [topology key](https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#interlude-built-in-node-labels)
   for pod anti affinity. The default is `kubernetes.io/hostname`.
 
+* **pod_management_policy**
+  specify the
+  [pod management policy](https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/#pod-management-policies)
+  of stateful sets of PG clusters. The default is `ordered_ready`, the second possible value is `parallel`.
+
 ## Kubernetes resource requests
 
 This group allows you to configure resource requests for the Postgres pods.
diff --git a/manifests/configmap.yaml b/manifests/configmap.yaml
@@ -44,6 +44,7 @@ data:
   pod_terminate_grace_period: 5m
   pod_deletion_wait_timeout: 10m
   pod_label_wait_timeout: 10m
+  pod_management_policy: "ordered_ready"
   ready_wait_interval: 3s
   ready_wait_timeout: 30s
   #  master_pod_move_timeout: 10m
diff --git a/pkg/cluster/k8sres.go b/pkg/cluster/k8sres.go
@@ -858,6 +858,20 @@ func (c *Cluster) generateStatefulSet(spec *acidv1.PostgresSpec) (*v1beta1.State
 
 	numberOfInstances := c.getNumberOfInstances(spec)
 
+	// the operator has domain-specific logic on how to do rolling updates of PG clusters
+	// so we do not use default rolling updates implemented by stateful sets
+	// that leaves the legacy "OnDelete" update strategy as the only option
+	updateStrategy := v1beta1.StatefulSetUpdateStrategy{Type: v1beta1.OnDeleteStatefulSetStrategyType}
+
+	var podManagementPolicy v1beta1.PodManagementPolicyType
+	if c.OpConfig.PodManagementPolicy == "ordered_ready" {
+		podManagementPolicy = v1beta1.OrderedReadyPodManagement
+	} else if c.OpConfig.PodManagementPolicy == "parallel" {
+		podManagementPolicy = v1beta1.ParallelPodManagement
+	} else {
+		return nil, fmt.Errorf("could not set the pod management policy to the unknown value: %v", c.OpConfig.PodManagementPolicy)
+	}
+
 	statefulSet := &v1beta1.StatefulSet{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:        c.statefulSetName(),
@@ -871,6 +885,8 @@ func (c *Cluster) generateStatefulSet(spec *acidv1.PostgresSpec) (*v1beta1.State
 			ServiceName:          c.serviceName(Master),
 			Template:             *podTemplate,
 			VolumeClaimTemplates: []v1.PersistentVolumeClaim{*volumeClaimTemplate},
+			UpdateStrategy:       updateStrategy,
+			PodManagementPolicy:  podManagementPolicy,
 		},
 	}
 
diff --git a/pkg/controller/node.go b/pkg/controller/node.go
@@ -184,7 +184,7 @@ func (c *Controller) moveMasterPodsOffNode(node *v1.Node) {
 	)
 
 	if err != nil {
-		c.logger.Warning("failed to move master pods from the node %q: timeout of %v minutes expired", node.Name, c.opConfig.MasterPodMoveTimeout)
+		c.logger.Warningf("failed to move master pods from the node %q: timeout of %v minutes expired", node.Name, c.opConfig.MasterPodMoveTimeout)
 	}
 
 }
diff --git a/pkg/util/config/config.go b/pkg/util/config/config.go
@@ -109,6 +109,7 @@ type Config struct {
 	ClusterHistoryEntries    int               `name:"cluster_history_entries" default:"1000"`
 	TeamAPIRoleConfiguration map[string]string `name:"team_api_role_configuration" default:"log_statement:all"`
 	PodTerminateGracePeriod  time.Duration     `name:"pod_terminate_grace_period" default:"5m"`
+	PodManagementPolicy      string            `name:"pod_management_policy" default:"ordered_ready"`
 	ProtectedRoles           []string          `name:"protected_role_names" default:"admin"`
 	PostgresSuperuserTeams   []string          `name:"postgres_superuser_teams" default:""`
 	SetMemoryRequestToLimit  bool              `name:"set_memory_request_to_limit" defaults:"false"`

Original file line number	Diff line number	Diff line change
`@@ -184,7 +184,7 @@ func (c Controller) moveMasterPodsOffNode(node v1.Node) {`
`184`	`184`	`)`
`185`	`185`
`186`	`186`	`if err != nil {`
`187`		`- c.logger.Warning("failed to move master pods from the node %q: timeout of %v minutes expired", node.Name, c.opConfig.MasterPodMoveTimeout)`
	`187`	`+ c.logger.Warningf("failed to move master pods from the node %q: timeout of %v minutes expired", node.Name, c.opConfig.MasterPodMoveTimeout)`
`188`	`188`	`}`
`189`	`189`
`190`	`190`	`}`