[Bugfix] Fix status propagation (#889)

CHANGELOG.md

@@ -16,6 +16,7 @@
 - Add ArangoClusterSynchronization Operator
 - Update licenses
 - Fix restart procedure in case of failing members
+- Fix status propagation race condition
 
 ## [1.2.6](https://github.com/arangodb/kube-arangodb/tree/1.2.6) (2021-12-15)
 - Add ArangoBackup backoff functionality

pkg/apis/deployment/v1/conditions.go

@@ -36,6 +36,10 @@ func (c ConditionType) String() string {
 const (
 	// ConditionTypeReady indicates that the member or entire deployment is ready and running normally.
 	ConditionTypeReady ConditionType = "Ready"
+	// ConditionTypeStarted indicates that the member was ready at least once.
+	ConditionTypeStarted ConditionType = "Started"
+	// ConditionTypeServing indicates that the member core services are running.
+	ConditionTypeServing ConditionType = "Serving"
 	// ConditionTypeTerminated indicates that the member has terminated and will not restart.
 	ConditionTypeTerminated ConditionType = "Terminated"
 	// ConditionTypeAutoUpgrade indicates that the member has to be started with `--database.auto-upgrade` once.

pkg/apis/deployment/v1/member_status_list.go

@@ -205,6 +205,22 @@ func (l MemberStatusList) MembersReady() int {
 	return readyCount
 }
 
+// MembersServing returns the number of members that are in the Serving state.
+func (l MemberStatusList) MembersServing() int {
+	servingCount := 0
+	for _, x := range l {
+		if x.Conditions.IsTrue(ConditionTypeServing) {
+			servingCount++
+		}
+	}
+	return servingCount
+}
+
+// AllMembersServing returns the true if all members are in the Serving state.
+func (l MemberStatusList) AllMembersServing() bool {
+	return len(l) == l.MembersServing()
+}
+
 // AllMembersReady returns the true if all members are in the Ready state.
 func (l MemberStatusList) AllMembersReady() bool {
 	return len(l) == l.MembersReady()

pkg/apis/deployment/v2alpha1/conditions.go

@@ -36,6 +36,10 @@ func (c ConditionType) String() string {
 const (
 	// ConditionTypeReady indicates that the member or entire deployment is ready and running normally.
 	ConditionTypeReady ConditionType = "Ready"
+	// ConditionTypeStarted indicates that the member was ready at least once.
+	ConditionTypeStarted ConditionType = "Started"
+	// ConditionTypeServing indicates that the member core services are running.
+	ConditionTypeServing ConditionType = "Serving"
 	// ConditionTypeTerminated indicates that the member has terminated and will not restart.
 	ConditionTypeTerminated ConditionType = "Terminated"
 	// ConditionTypeAutoUpgrade indicates that the member has to be started with `--database.auto-upgrade` once.

pkg/apis/deployment/v2alpha1/member_status_list.go

@@ -205,6 +205,22 @@ func (l MemberStatusList) MembersReady() int {
 	return readyCount
 }
 
+// MembersServing returns the number of members that are in the Serving state.
+func (l MemberStatusList) MembersServing() int {
+	servingCount := 0
+	for _, x := range l {
+		if x.Conditions.IsTrue(ConditionTypeServing) {
+			servingCount++
+		}
+	}
+	return servingCount
+}
+
+// AllMembersServing returns the true if all members are in the Serving state.
+func (l MemberStatusList) AllMembersServing() bool {
+	return len(l) == l.MembersServing()
+}
+
 // AllMembersReady returns the true if all members are in the Ready state.
 func (l MemberStatusList) AllMembersReady() bool {
 	return len(l) == l.MembersReady()

pkg/deployment/deployment.go

@@ -52,7 +52,7 @@ import (
 	"github.com/arangodb/arangosync-client/client"
 	"github.com/rs/zerolog"
 	apiextensionsclient "k8s.io/apiextensions-apiserver/pkg/client/clientset/clientset"
-	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	meta "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/client-go/kubernetes"
 	"k8s.io/client-go/tools/record"
 
@@ -381,7 +381,7 @@ func (d *Deployment) handleArangoDeploymentUpdatedEvent(ctx context.Context) err
 	// Get the most recent version of the deployment from the API server
 	ctxChild, cancel := globals.GetGlobalTimeouts().Kubernetes().WithTimeout(ctx)
 	defer cancel()
-	current, err := d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(d.apiObject.GetNamespace()).Get(ctxChild, d.apiObject.GetName(), metav1.GetOptions{})
+	current, err := d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(d.apiObject.GetNamespace()).Get(ctxChild, d.apiObject.GetName(), meta.GetOptions{})
 	if err != nil {
 		log.Debug().Err(err).Msg("Failed to get current version of deployment from API server")
 		if k8sutil.IsNotFound(err) {
@@ -465,21 +465,22 @@ func (d *Deployment) updateCRStatus(ctx context.Context, force ...bool) error {
 	}
 
 	// Send update to API server
-	ns := d.apiObject.GetNamespace()
-	depls := d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(ns)
-	update := d.apiObject.DeepCopy()
+	depls := d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(d.GetNamespace())
 	attempt := 0
 	for {
 		attempt++
-		update.Status = d.status.last
-		if update.GetDeletionTimestamp() == nil {
-			ensureFinalizers(update)
+		if d.apiObject.GetDeletionTimestamp() == nil {
+			ensureFinalizers(d.apiObject)
 		}
 
 		var newAPIObject *api.ArangoDeployment
 		err := globals.GetGlobalTimeouts().Kubernetes().RunWithTimeout(ctx, func(ctxChild context.Context) error {
-			var err error
-			newAPIObject, err = depls.Update(ctxChild, update, metav1.UpdateOptions{})
+			p, err := patch.NewPatch(patch.ItemReplace(patch.NewPath("status"), d.status.last)).Marshal()
+			if err != nil {
+				return err
+			}
+
+			newAPIObject, err = depls.Patch(ctxChild, d.GetName(), types.JSONPatchType, p, meta.PatchOptions{})
 
 			return err
 		})
@@ -488,21 +489,8 @@ func (d *Deployment) updateCRStatus(ctx context.Context, force ...bool) error {
 			d.apiObject = newAPIObject
 			return nil
 		}
-		if attempt < 10 && k8sutil.IsConflict(err) {
-			// API object may have been changed already,
-			// Reload api object and try again
-			var current *api.ArangoDeployment
-
-			err = globals.GetGlobalTimeouts().Kubernetes().RunWithTimeout(ctx, func(ctxChild context.Context) error {
-				var err error
-				current, err = depls.Get(ctxChild, update.GetName(), metav1.GetOptions{})
-
-				return err
-			})
-			if err == nil {
-				update = current.DeepCopy()
-				continue
-			}
+		if attempt < 10 {
+			continue
 		}
 		if err != nil {
 			d.deps.Log.Debug().Err(err).Msg("failed to patch ArangoDeployment status")
@@ -535,7 +523,7 @@ func (d *Deployment) updateCRSpec(ctx context.Context, newSpec api.DeploymentSpe
 		var newAPIObject *api.ArangoDeployment
 		err := globals.GetGlobalTimeouts().Kubernetes().RunWithTimeout(ctx, func(ctxChild context.Context) error {
 			var err error
-			newAPIObject, err = d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(ns).Update(ctxChild, update, metav1.UpdateOptions{})
+			newAPIObject, err = d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(ns).Update(ctxChild, update, meta.UpdateOptions{})
 
 			return err
 		})
@@ -551,7 +539,7 @@ func (d *Deployment) updateCRSpec(ctx context.Context, newSpec api.DeploymentSpe
 
 			err = globals.GetGlobalTimeouts().Kubernetes().RunWithTimeout(ctx, func(ctxChild context.Context) error {
 				var err error
-				current, err = d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(ns).Get(ctxChild, update.GetName(), metav1.GetOptions{})
+				current, err = d.deps.DatabaseCRCli.DatabaseV1().ArangoDeployments(ns).Get(ctxChild, update.GetName(), meta.GetOptions{})
 
 				return err
 			})
@@ -568,7 +556,7 @@ func (d *Deployment) updateCRSpec(ctx context.Context, newSpec api.DeploymentSpe
 }
 
 // isOwnerOf returns true if the given object belong to this deployment.
-func (d *Deployment) isOwnerOf(obj metav1.Object) bool {
+func (d *Deployment) isOwnerOf(obj meta.Object) bool {
 	ownerRefs := obj.GetOwnerReferences()
 	if len(ownerRefs) < 1 {
 		return false
@@ -583,9 +571,9 @@ func (d *Deployment) isOwnerOf(obj metav1.Object) bool {
 func (d *Deployment) lookForServiceMonitorCRD() {
 	var err error
 	if d.GetScope().IsNamespaced() {
-		_, err = d.deps.KubeMonitoringCli.ServiceMonitors(d.GetNamespace()).List(context.Background(), metav1.ListOptions{})
+		_, err = d.deps.KubeMonitoringCli.ServiceMonitors(d.GetNamespace()).List(context.Background(), meta.ListOptions{})
 	} else {
-		_, err = d.deps.KubeExtCli.ApiextensionsV1().CustomResourceDefinitions().Get(context.Background(), "servicemonitors.monitoring.coreos.com", metav1.GetOptions{})
+		_, err = d.deps.KubeExtCli.ApiextensionsV1().CustomResourceDefinitions().Get(context.Background(), "servicemonitors.monitoring.coreos.com", meta.GetOptions{})
 	}
 	log := d.deps.Log
 	log.Debug().Msgf("Looking for ServiceMonitor CRD...")
@@ -637,7 +625,7 @@ func (d *Deployment) ApplyPatch(ctx context.Context, p ...patch.Item) error {
 
 	ctxChild, cancel := globals.GetGlobalTimeouts().Kubernetes().WithTimeout(ctx)
 	defer cancel()
-	depl, err := c.Patch(ctxChild, d.apiObject.GetName(), types.JSONPatchType, data, metav1.PatchOptions{})
+	depl, err := c.Patch(ctxChild, d.apiObject.GetName(), types.JSONPatchType, data, meta.PatchOptions{})
 	if err != nil {
 		return err
 	}

pkg/deployment/member/phase_updates.go

@@ -67,6 +67,7 @@ var phase = phaseMap{
 func removeMemberConditionsMapFunc(m *api.MemberStatus) {
 	// Clean conditions
 	m.Conditions.Remove(api.ConditionTypeReady)
+	m.Conditions.Remove(api.ConditionTypeStarted)
 	m.Conditions.Remove(api.ConditionTypeTerminated)
 	m.Conditions.Remove(api.ConditionTypeTerminating)
 	m.Conditions.Remove(api.ConditionTypeAgentRecoveryNeeded)

pkg/deployment/reconcile/plan_builder_rotate_upgrade.go

@@ -384,18 +384,23 @@ func groupReadyForRestart(context PlanBuilderContext, spec api.DeploymentSpec, s
 		return true
 	}
 
-	// If current member is not ready, kill anyway
-	if !member.Conditions.IsTrue(api.ConditionTypeReady) {
+	// If current member did not become ready even once. Kill it
+	if !member.Conditions.IsTrue(api.ConditionTypeStarted) {
+		return true
+	}
+
+	// If current core containers are dead kill it.
+	if !member.Conditions.IsTrue(api.ConditionTypeServing) {
 		return true
 	}
 
 	switch group {
 	case api.ServerGroupDBServers:
 		// TODO: Improve shard placement discovery and keep WriteConcern
-		return context.GetShardSyncStatus() && status.Members.MembersOfGroup(group).AllMembersReady()
+		return context.GetShardSyncStatus() && status.Members.MembersOfGroup(group).AllMembersServing()
 	default:
 		// In case of agents we can kill only one agent at same time
-		return status.Members.MembersOfGroup(group).AllMembersReady()
+		return status.Members.MembersOfGroup(group).AllMembersServing()
 	}
 }
 

pkg/deployment/reconcile/plan_builder_utils.go

@@ -60,8 +60,8 @@ func emptyPlanBuilder(ctx context.Context,
 
 func removeConditionActionV2(actionReason string, conditionType api.ConditionType) api.Action {
 	return api.NewAction(api.ActionTypeSetConditionV2, api.ServerGroupUnknown, "", actionReason).
-		AddParam(setConditionActionV2KeyAction, setConditionActionV2KeyTypeRemove).
-		AddParam(setConditionActionV2KeyType, string(conditionType))
+		AddParam(setConditionActionV2KeyAction, string(conditionType)).
+		AddParam(setConditionActionV2KeyType, setConditionActionV2KeyTypeRemove)
 }
 
 func updateConditionActionV2(actionReason string, conditionType api.ConditionType, status bool, reason, message, hash string) api.Action {

pkg/deployment/resources/pod_inspector.go

@@ -35,6 +35,8 @@ import (
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
+	"strings"
+
 	api "github.com/arangodb/kube-arangodb/pkg/apis/deployment/v1"
 	"github.com/arangodb/kube-arangodb/pkg/metrics"
 	"github.com/arangodb/kube-arangodb/pkg/util"
@@ -216,10 +218,12 @@ func (r *Resources) InspectPods(ctx context.Context, cachedStatus inspectorInter
 		}
 		// End of Topology labels
 
-		if k8sutil.AreContainersReady(pod, coreContainers) {
+		if k8sutil.IsPodReady(pod) && k8sutil.AreContainersReady(pod, coreContainers) {
 			// Pod is now ready
-			if memberStatus.Conditions.Update(api.ConditionTypeReady, true, "Pod Ready", "") {
-				log.Debug().Str("pod-name", pod.GetName()).Msg("Updating member condition Ready & Initialised to true")
+			if anyOf(memberStatus.Conditions.Update(api.ConditionTypeReady, true, "Pod Ready", ""),
+				memberStatus.Conditions.Update(api.ConditionTypeStarted, true, "Pod Started", ""),
+				memberStatus.Conditions.Update(api.ConditionTypeServing, true, "Pod Serving", "")) {
+				log.Debug().Str("pod-name", pod.GetName()).Msg("Updating member condition Ready, Started & Serving to true")
 
 				if status.Topology.IsTopologyOwned(memberStatus.Topology) {
 					nodes, ok := cachedStatus.GetNodes()
@@ -238,10 +242,19 @@ func (r *Resources) InspectPods(ctx context.Context, cachedStatus inspectorInter
 				updateMemberStatusNeeded = true
 				nextInterval = nextInterval.ReduceTo(recheckSoonPodInspectorInterval)
 			}
+		} else if k8sutil.AreContainersReady(pod, coreContainers) {
+			// Pod is not ready, but core containers are fine
+			if anyOf(memberStatus.Conditions.Update(api.ConditionTypeReady, false, "Pod Not Ready", ""),
+				memberStatus.Conditions.Update(api.ConditionTypeServing, true, "Pod is still serving", "")) {
+				log.Debug().Str("pod-name", pod.GetName()).Msg("Updating member condition Ready to false, while all core containers are ready")
+				updateMemberStatusNeeded = true
+				nextInterval = nextInterval.ReduceTo(recheckSoonPodInspectorInterval)
+			}
 		} else {
 			// Pod is not ready
-			if memberStatus.Conditions.Update(api.ConditionTypeReady, false, "Pod Not Ready", "") {
-				log.Debug().Str("pod-name", pod.GetName()).Msg("Updating member condition Ready to false")
+			if anyOf(memberStatus.Conditions.Update(api.ConditionTypeReady, false, "Pod Not Ready", ""),
+				memberStatus.Conditions.Update(api.ConditionTypeServing, false, "Pod Core containers are not ready", strings.Join(coreContainers, ", "))) {
+				log.Debug().Str("pod-name", pod.GetName()).Msg("Updating member condition Ready & Serving to false")
 				updateMemberStatusNeeded = true
 				nextInterval = nextInterval.ReduceTo(recheckSoonPodInspectorInterval)
 			}
@@ -394,3 +407,13 @@ func removeLabel(labels map[string]string, key string) map[string]string {
 
 	return labels
 }
+
+func anyOf(bools ...bool) bool {
+	for _, b := range bools {
+		if b {
+			return true
+		}
+	}
+
+	return false
+}

pkg/util/k8sutil/pods.go

@@ -120,13 +120,14 @@ func AreContainersReady(pod *core.Pod, coreContainers utils.StringList) bool {
 	// From here on all required containers are running, but unready condition must be checked additionally.
 	switch condition.Reason {
 	case ServerContainerConditionContainersNotReady:
-		if !strings.HasPrefix(condition.Message, ServerContainerConditionPrefix) {
+		unreadyContainers, ok := extractContainerNamesFromConditionMessage(condition.Message)
+
+		if !ok {
 			return false
 		}
 
-		unreadyContainers := strings.TrimPrefix(condition.Message, ServerContainerConditionPrefix)
 		for _, c := range coreContainers {
-			if strings.Contains(unreadyContainers, c) {
+			if unreadyContainers.Has(c) {
 				// The container is on the list with unready containers.
 				return false
 			}
@@ -138,6 +139,28 @@ func AreContainersReady(pod *core.Pod, coreContainers utils.StringList) bool {
 	return false
 }
 
+func extractContainerNamesFromConditionMessage(msg string) (utils.StringList, bool) {
+	if !strings.HasPrefix(msg, ServerContainerConditionPrefix) {
+		return nil, false
+	}
+
+	unreadyContainers := strings.TrimPrefix(msg, ServerContainerConditionPrefix)
+
+	if !strings.HasPrefix(unreadyContainers, "[") {
+		return nil, false
+	}
+
+	if !strings.HasSuffix(unreadyContainers, "]") {
+		return nil, false
+	}
+
+	unreadyContainers = strings.TrimPrefix(strings.TrimSuffix(unreadyContainers, "]"), "[")
+
+	unreadyContainersList := utils.StringList(strings.Split(unreadyContainers, " "))
+
+	return unreadyContainersList, true
+}
+
 // GetPodByName returns pod if it exists among the pods' list
 // Returns false if not found.
 func GetPodByName(pods []core.Pod, podName string) (core.Pod, bool) {

pkg/util/k8sutil/pods_test.go

@@ -27,6 +27,7 @@ import (
 	v1 "k8s.io/api/core/v1"
 
 	"github.com/arangodb/kube-arangodb/pkg/handlers/utils"
+	"github.com/stretchr/testify/require"
 )
 
 // TestIsPodReady tests IsPodReady.
@@ -318,3 +319,14 @@ func TestIsPodSucceeded(t *testing.T) {
 		})
 	}
 }
+
+func Test_extractContainerNamesFromConditionMessage(t *testing.T) {
+	t.Run("Valid name", func(t *testing.T) {
+		c, ok := extractContainerNamesFromConditionMessage("containers with unready status: [sidecar2 sidecar3]")
+		require.True(t, ok)
+		require.Len(t, c, 2)
+		require.Contains(t, c, "sidecar2")
+		require.Contains(t, c, "sidecar3")
+		require.NotContains(t, c, "sidecar")
+	})
+}