[Feature] Agency DBServer Discovery (#1820)

Author: ajanikow

CHANGELOG.md

@@ -9,6 +9,7 @@
 - (Feature) (Platform) CLI
 - (Bugfix) Fix Condition name
 - (Bugfix) Add missing ArangoDeployment ExternalAccess Managed Type definition
+- (Feature) Agency DBServer Discovery
 
 ## [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/state/state.go

@@ -20,8 +20,6 @@
 
 package state
 
-import "github.com/arangodb/go-driver"
-
 type Root struct {
 	Arango   State `json:"arango"`
 	ArangoDB DB    `json:"arangodb,omitempty"`
@@ -53,12 +51,14 @@ type Current struct {
 	Collections        CurrentCollections        `json:"Collections"`
 
 	// ServersKnown stores information about ArangoDB servers.
-	ServersKnown map[driver.ServerID]ServerKnown `json:"ServersKnown,omitempty"`
+	ServersKnown ServerMap[ServerKnown] `json:"ServersKnown,omitempty"`
 }
 
 type Plan struct {
-	Collections PlanCollections `json:"Collections"`
-	Databases   PlanDatabases   `json:"Databases,omitempty"`
+	Collections  PlanCollections   `json:"Collections"`
+	Databases    PlanDatabases     `json:"Databases,omitempty"`
+	DBServers    ServerMap[string] `json:"DBServers,omitempty"`
+	Coordinators ServerMap[string] `json:"Coordinators,omitempty"`
 }
 
 type Supervision struct {
@@ -69,6 +69,13 @@ type ShardCountDetails struct {
 	Leader, Follower int
 }
 
+type ServerMap[T any] map[Server]T
+
+func (s ServerMap[T]) Exists(server Server) bool {
+	_, ok := s[server]
+	return ok
+}
+
 func (s ShardCountDetails) Count() int {
 	return s.Leader + s.Follower
 }
@@ -445,7 +452,7 @@ func (s State) GetCollectionDatabaseByID(id string) (string, bool) {
 
 // GetRebootID returns reboot ID for a given server ID.
 // returns false when a server ID does not exist in cache.
-func (s State) GetRebootID(id driver.ServerID) (int, bool) {
+func (s State) GetRebootID(id Server) (int, bool) {
 	if v, ok := s.Current.ServersKnown[id]; ok {
 		return v.RebootID, true
 	}

pkg/deployment/agency/state/state_test.go

@@ -70,6 +70,7 @@ var (
 		"3.9":           agencyDump39,
 		"3.9-satellite": agencyDump39Satellite,
 		"3.9-hotbackup": agencyDump39HotBackup,
+		"3.10":          agencyDump310,
 	}
 )
 
@@ -87,6 +88,11 @@ func Test_Unmarshal_MultiVersion(t *testing.T) {
 				}
 			})
 
+			t.Run("Ensure member discovery", func(t *testing.T) {
+				require.NotEmpty(t, s.Agency.Arango.Plan.DBServers)
+				require.NotEmpty(t, s.Agency.Arango.Plan.Coordinators)
+			})
+
 			t.Run("Ensure distributeShardsLike", func(t *testing.T) {
 				collections, ok := s.Agency.Arango.Plan.Collections["_system"]
 				require.True(t, ok)

pkg/deployment/reconcile/action_enforce_resign_leadership.go

@@ -1,7 +1,7 @@
 //
 // DISCLAIMER
 //
-// Copyright 2023 ArangoDB GmbH, Cologne, Germany
+// Copyright 2023-2025 ArangoDB GmbH, Cologne, Germany
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -98,7 +98,7 @@ func (a *actionEnforceResignLeadership) CheckProgress(ctx context.Context) (bool
 		a.log.Warn("Maintenance is enabled, skipping action")
 		// We are done, action cannot be handled on maintenance mode
 		return true, false, nil
-	} else if isServerRebooted(a.log, a.action, agencyState, driver.ServerID(m.ID)) {
+	} else if hasServerRebooted(a.log, a.action, agencyState, state.Server(m.ID)) {
 		return true, false, nil
 	}
 

pkg/deployment/reconcile/action_resign_leadership.go

@@ -1,7 +1,7 @@
 //
 // DISCLAIMER
 //
-// Copyright 2016-2023 ArangoDB GmbH, Cologne, Germany
+// Copyright 2016-2025 ArangoDB GmbH, Cologne, Germany
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -131,7 +131,7 @@ func (a *actionResignLeadership) CheckProgress(ctx context.Context) (bool, bool,
 			return false, false, errors.WithStack(err)
 		}
 		return true, false, nil
-	} else if isServerRebooted(a.log, a.action, agencyState, driver.ServerID(m.ID)) {
+	} else if hasServerRebooted(a.log, a.action, agencyState, state.Server(m.ID)) {
 		return true, false, nil
 	}
 

pkg/deployment/reconcile/action_resign_leadership_utils.go

@@ -1,7 +1,7 @@
 //
 // DISCLAIMER
 //
-// Copyright 2023 ArangoDB GmbH, Cologne, Germany
+// Copyright 2023-2025 ArangoDB GmbH, Cologne, Germany
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -23,8 +23,6 @@ package reconcile
 import (
 	"strconv"
 
-	"github.com/arangodb/go-driver"
-
 	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/deployment/features"
@@ -38,8 +36,8 @@ func getResignLeadershipActionType() api.ActionType {
 	return api.ActionTypeResignLeadership
 }
 
-// isServerRebooted returns true when a given server ID was rebooted during resignation of leadership.
-func isServerRebooted(log logging.Logger, action api.Action, agencyState state.State, serverID driver.ServerID) bool {
+// hasServerRebooted returns true when a given server ID was rebooted during resignation of leadership.
+func hasServerRebooted(log logging.Logger, action api.Action, agencyState state.State, serverID state.Server) bool {
 	rebootID, ok := agencyState.GetRebootID(serverID)
 	if !ok {
 		return false

pkg/deployment/reconcile/action_wait_for_member_ready.go

@@ -1,7 +1,7 @@
 //
 // DISCLAIMER
 //
-// Copyright 2022 ArangoDB GmbH, Cologne, Germany
+// Copyright 2022-2025 ArangoDB GmbH, Cologne, Germany
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -24,6 +24,7 @@ import (
 	"context"
 
 	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"
 )
 
@@ -67,5 +68,24 @@ func (a *actionWaitForMemberReady) CheckProgress(ctx context.Context) (bool, boo
 		return true, false, nil
 	}
 
+	cache, ok := a.actionCtx.GetAgencyCache()
+	if !ok {
+		a.log.Debug("AgencyCache is not ready")
+		return false, false, nil
+	}
+
+	switch a.action.Group {
+	case api.ServerGroupDBServers:
+		if !cache.Plan.DBServers.Exists(state.Server(member.ID)) {
+			a.log.Debug("DBServer not yet present")
+			return false, false, nil
+		}
+	case api.ServerGroupCoordinators:
+		if !cache.Plan.Coordinators.Exists(state.Server(member.ID)) {
+			a.log.Debug("Coordinator not yet present")
+			return false, false, nil
+		}
+	}
+
 	return member.Conditions.IsTrue(api.ConditionTypeReady), false, nil
 }

pkg/deployment/reconcile/plan_builder_rotate_upgrade.go

@@ -196,7 +196,7 @@ func (r *Reconciler) createUpdatePlanInternal(apiObject k8sutil.APIObject, spec
 		}
 
 		if m.Member.Conditions.IsTrue(api.ConditionTypeRestart) {
-			return r.createRotateMemberPlan(m.Member, m.Group, spec, "Restart flag present", util.CheckConditionalP1Nil(agencyCache.GetRebootID, driver.ServerID(m.Member.ID))), false
+			return r.createRotateMemberPlan(m.Member, m.Group, spec, "Restart flag present", util.CheckConditionalP1Nil(agencyCache.GetRebootID, state.Server(m.Member.ID))), false
 		}
 
 		arangoMember, ok := context.ACS().CurrentClusterCache().ArangoMember().V1().GetSimple(m.Member.ArangoMemberName(apiObject.GetName(), m.Group))
@@ -560,7 +560,7 @@ func (r *Reconciler) createUpgradeMemberPlan(member api.MemberStatus,
 		Str("action", string(upgradeAction)).
 		Info("Creating upgrade plan")
 
-	plan := createRotateMemberPlanWithAction(member, group, upgradeAction, spec, reason, util.CheckConditionalP1Nil(agencyCache.GetRebootID, driver.ServerID(member.ID)))
+	plan := createRotateMemberPlanWithAction(member, group, upgradeAction, spec, reason, util.CheckConditionalP1Nil(agencyCache.GetRebootID, state.Server(member.ID)))
 
 	if member.Image == nil || member.Image.Image != spec.GetImage() {
 		plan = plan.Before(actions.NewAction(api.ActionTypeSetMemberCurrentImage, group, member, reason).SetImage(spec.GetImage()))