[Bugfix] Agency Cache Reload

Author: ajanikow

CHANGELOG.md

@@ -11,6 +11,7 @@
 - (Bugfix) Add missing ArangoDeployment ExternalAccess Managed Type definition
 - (Feature) Agency DBServer Discovery
 - (Bugfix) Fix Manifests
+- (Bugfix) Agency Cache Reload
 
 ## [1.2.44](https://github.com/arangodb/kube-arangodb/tree/1.2.44) (2025-02-03)
 - (Maintenance) Kubernetes 1.31.1 libraries

pkg/deployment/agency/cache.go

@@ -155,6 +155,8 @@ type Cache interface {
 	Health() (Health, bool)
 	// ShardsInSyncMap returns last in sync state of shards. If no state is available, false is returned.
 	ShardsInSyncMap() (state.ShardsSyncStatus, bool)
+
+	Invalidate()
 }
 
 func NewCache(namespace, name string, mode *api.DeploymentMode) Cache {
@@ -214,6 +216,9 @@ func (c cacheSingle) Data() (state.State, bool) {
 	return state.State{}, true
 }
 
+func (c cacheSingle) Invalidate() {
+}
+
 type cache struct {
 	namespace, name string
 
@@ -228,6 +233,10 @@ type cache struct {
 	shardsSyncStatus state.ShardsSyncStatus
 }
 
+func (c *cache) Invalidate() {
+	c.loader.Invalidate()
+}
+
 func (c *cache) WrapLogger(in *zerolog.Event) *zerolog.Event {
 	return in.Str("namespace", c.namespace).Str("name", c.name)
 }

pkg/deployment/deployment.go

@@ -200,7 +200,12 @@ func (d *Deployment) RefreshAgencyCache(ctx context.Context) (uint64, error) {
 				clients[m.ID] = a
 			}
 
-			return d.agencyCache.Reload(ctx, rsize, clients)
+			if offset, err := d.agencyCache.Reload(ctx, rsize, clients); err != nil {
+				d.agencyCache.Invalidate()
+				return 0, err
+			} else {
+				return offset, nil
+			}
 		}
 	}
 

pkg/deployment/reconcile/action_enforce_resign_leadership.go

@@ -27,7 +27,6 @@ import (
 
 	api "github.com/arangodb/kube-arangodb/pkg/apis/deployment/v1"
 	"github.com/arangodb/kube-arangodb/pkg/deployment/agency/state"
-	"github.com/arangodb/kube-arangodb/pkg/util/errors"
 	"github.com/arangodb/kube-arangodb/pkg/util/globals"
 )
 
@@ -142,15 +141,15 @@ func (a *actionEnforceResignLeadership) CheckProgress(ctx context.Context) (bool
 	client, err := a.actionCtx.GetMembersState().State().GetDatabaseClient()
 	if err != nil {
 		a.log.Err(err).Error("Unable to get client")
-		return false, false, errors.WithStack(err)
+		return false, false, nil
 	}
 
 	ctxChild, cancel := globals.GetGlobalTimeouts().ArangoD().WithTimeout(ctx)
 	defer cancel()
 	cluster, err := client.Cluster(ctxChild)
 	if err != nil {
 		a.log.Err(err).Error("Unable to get cluster client")
-		return false, false, errors.WithStack(err)
+		return false, false, nil
 	}
 
 	var jobID string
@@ -160,7 +159,7 @@ func (a *actionEnforceResignLeadership) CheckProgress(ctx context.Context) (bool
 	a.log.Debug("Temporary shutdown, resign leadership")
 	if err := cluster.ResignServer(jobCtx, m.ID); err != nil {
 		a.log.Err(err).Debug("Failed to resign server")
-		return false, false, errors.WithStack(err)
+		return false, false, nil
 	}
 
 	a.actionCtx.Add(resignLeadershipJobID, jobID, true)