Switch IPPool agent to Deployment with rolling updates

Author: davidepasquero1

chart/crds/network.harvesterhci.io_ippools.yaml

@@ -150,7 +150,7 @@ spec:
             type: object
           status:
             properties:
-              agentPodRef:
+              agentDeploymentRef:
                 properties:
                   image:
                     type: string

chart/templates/rbac.yaml

@@ -18,9 +18,9 @@ rules:
 - apiGroups: [ "" ]
   resources: [ "namespaces" ]
   verbs: [ "get", "watch", "list" ]
-- apiGroups: [ "" ]
-  resources: [ "pods" ]
-  verbs: [ "watch", "list" ]
+- apiGroups: [ "apps" ]
+  resources: [ "deployments" ]
+  verbs: [ "get", "watch", "list" ]
 - apiGroups: [ "kubevirt.io" ]
   resources: [ "virtualmachines" ]
   verbs: [ "get", "watch", "list" ]
@@ -122,11 +122,11 @@ rules:
 apiVersion: rbac.authorization.k8s.io/v1
 kind: Role
 metadata:
-  name: {{ include "harvester-vm-dhcp-controller.name" . }}-pod-manager
+  name: {{ include "harvester-vm-dhcp-controller.name" . }}-deployment-manager
 rules:
-- apiGroups: [ "" ]
-  resources: [ "pods" ]
-  verbs: [ "get", "create", "delete" ]
+- apiGroups: [ "apps" ]
+  resources: [ "deployments" ]
+  verbs: [ "get", "create", "update", "delete" ]
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: Role
@@ -157,13 +157,13 @@ subjects:
 apiVersion: rbac.authorization.k8s.io/v1
 kind: RoleBinding
 metadata:
-  name: {{ include "harvester-vm-dhcp-controller.name" . }}-manage-pods
+  name: {{ include "harvester-vm-dhcp-controller.name" . }}-manage-deployments
   labels:
   {{- include "harvester-vm-dhcp-controller.labels" . | nindent 4 }}
 roleRef:
   apiGroup: rbac.authorization.k8s.io
   kind: Role
-  name: {{ include "harvester-vm-dhcp-controller.name" . }}-pod-manager
+  name: {{ include "harvester-vm-dhcp-controller.name" . }}-deployment-manager
 subjects:
 - kind: ServiceAccount
   name: {{ include "harvester-vm-dhcp-controller.serviceAccountName" . }}

docs/ippool-agent-deployment.md

@@ -0,0 +1,129 @@
+---
+name: ippool-agent-deployment-reconcile
+description: Plan to move IPPool agents from Pods to Deployments and reconcile via operator pattern.
+---
+
+# Plan
+
+Obiettivo: passare da agent Pod a Deployment per ogni IPPool e riconciliare con pattern operator, aggiornando API, controller, RBAC e test.
+
+## Requirements
+- Ogni IPPool crea/gestisce un Deployment agent (repliche=1) con la stessa logica di rete/affinity/args attuale.
+- Reconcile idempotente: crea/aggiorna/monitor/cleanup del Deployment.
+- Upgrade immagine rispettando `hold-ippool-agent-upgrade`.
+
+## Scope
+- In: IPPool controller, API/CRD status, codegen, RBAC, test.
+- Out: logica DHCP/IPAM/VMNetCfg non correlata.
+
+## Files and entry points
+- `pkg/controller/ippool/controller.go`
+- `pkg/controller/ippool/common.go`
+- `pkg/apis/network.harvesterhci.io/v1alpha1/ippool.go`
+- `pkg/codegen/main.go`, `pkg/config/context.go`
+- `pkg/util/fakeclient`, `pkg/controller/ippool/controller_test.go`
+- `chart/templates/rbac.yaml`, `chart/crds/network.harvesterhci.io_ippools.yaml`
+- `pkg/data/data.go`
+
+## Data model / API changes
+- Sostituire `status.agentPodRef` con `status.agentDeploymentRef` (nuovo `DeploymentReference`).
+
+## Action items
+[ ] Aggiungere client/controller apps/v1 Deployment (codegen + Management) e fakeclient.
+[ ] Implementare `prepareAgentDeployment` e aggiornare Deploy/Monitor/Cleanup per Deployment.
+[ ] Aggiornare watch `relatedresource` su Deployment con label ippool.
+[ ] Aggiornare builder/status helpers e test per Deployment.
+[ ] Aggiornare RBAC per `deployments` (get/list/watch/create/update/delete).
+[ ] Rigenerare codegen/CRD/bindata.
+
+## Testing and validation
+- `go test ./...` (o `go test ./pkg/controller/ippool -run TestHandler_`).
+- `go generate` per rigenerare CRD/client/bindata.
+
+## Risks and edge cases
+- Breaking change CRD status.
+- Strategia Deployment (Recreate vs RollingUpdate) impatta continuita DHCP.
+- Namespace agent != IPPool: relazione via label, non ownerRef.
+
+## Open questions
+- Posso usare `agentDeploymentRef` al posto di `agentPodRef` anche se breaking?
+- Preferisci `Recreate` o `RollingUpdate` per i Deployment agent?
+
+---
+
+# IPPool agent su Deployment
+
+Questo documento descrive logica e modifiche introdotte per spostare l'agent IPPool
+da Pod singolo a Deployment, con riconciliazione di tipo operator e strategia
+RollingUpdate.
+
+## Obiettivo
+- Ogni IPPool crea e gestisce un Deployment dedicato (repliche=1).
+- Riconciliazione idempotente che crea o aggiorna il Deployment quando cambia la configurazione.
+- Upgrade immagine controllato dall'annotazione `network.harvesterhci.io/hold-ippool-agent-upgrade`.
+- Migrazione centrata sul nuovo riferimento di stato `agentDeploymentRef`.
+
+## Flusso di riconciliazione
+### DeployAgent
+- Recupera ClusterNetwork dalla NetworkAttachmentDefinition (label `network.harvesterhci.io/clusternetwork`).
+- Calcola l'immagine desiderata (rispetta l'annotazione di hold).
+- Costruisce il Deployment desiderato via `prepareAgentDeployment`.
+- Se `status.agentDeploymentRef` e valorizzato:
+  - Verifica UID e che il Deployment non sia in deletion.
+  - Richiede selector immutabile (errore se diverge).
+  - Aggiorna labels, strategy, replicas, template e container se differiscono.
+  - Aggiorna `agentDeploymentRef` con namespace/nome/UID.
+- Se il Deployment non esiste o lo status e vuoto, crea il Deployment e registra lo status.
+
+### MonitorAgent
+- Se `noAgent` e true, non fa nulla; se IPPool e in pausa, ritorna errore.
+- Verifica esistenza, UID e immagine del container rispetto allo status.
+- Verifica readiness usando `ObservedGeneration` e repliche `Updated/Available`.
+- In caso di mismatch o non readiness, ritorna errore (non cancella il Deployment).
+
+### Cleanup
+- Elimina il Deployment associato e pulisce IPAM/MAC/metriche.
+- Usato su pause o rimozione IPPool.
+
+## Dettagli implementativi
+- Nome Deployment derivato da `util.SafeAgentConcatName`.
+- Labels: `network.harvesterhci.io/vm-dhcp-controller=agent` + labels IPPool namespace/nome.
+- Pod template con annotazione Multus `k8s.v1.cni.cncf.io/networks`.
+- Init container `ip-setter` per configurare `eth1`; container `agent` con probes `/healthz` e `/readyz`.
+- Container defaults (ImagePullPolicy e TerminationMessage*) esplicitati per evitare reconcile loop.
+- Strategia RollingUpdate: `maxUnavailable=0`, `maxSurge=1`.
+
+## Cambiamenti di stato/CRD
+- `status.agentPodRef` sostituito da `status.agentDeploymentRef`.
+- Nuova struct `DeploymentReference` con namespace, name, image, UID.
+- CRD aggiornata e bindata rigenerata.
+
+## Watch e relazione risorse
+- Watch su Deployment con label `vm-dhcp-controller=agent`.
+- Mapping IPPool tramite label `ippool-namespace` e `ippool-name`.
+
+## RBAC
+- Aggiunti permessi `deployments` per controller (get/list/watch).
+- Role dedicato `*-deployment-manager` con create/update/delete.
+- Binding aggiornato da `manage-pods` a `manage-deployments`.
+
+## Codegen e generated
+- Aggiunto gruppo apps/v1 alla codegen per controller/cache Deployment.
+- Nuovi controller generati in `pkg/generated/controllers/apps`.
+- Nuovo fake client per Deployment in `pkg/util/fakeclient/deployment.go`.
+
+## Modifiche al codice (file principali)
+- `pkg/controller/ippool/controller.go`: DeployAgent/MonitorAgent/cleanup aggiornati per Deployment.
+- `pkg/controller/ippool/common.go`: `prepareAgentDeployment` e builder di supporto.
+- `pkg/apis/network.harvesterhci.io/v1alpha1/ippool.go`: nuovo status field e tipo reference.
+- `pkg/config/context.go`: AppsFactory aggiunta al bootstrap.
+- `pkg/codegen/main.go`: codegen apps/v1 e rigenerazione dei controller.
+- `chart/templates/rbac.yaml` e `chart/crds/network.harvesterhci.io_ippools.yaml`: RBAC/CRD aggiornate.
+
+## Compatibilita e migrazione
+- `status.agentPodRef` non e piu letto; eventuali Pod legacy non sono piu gestiti.
+- RollingUpdate con `maxSurge=1` puo generare un secondo agent temporaneo.
+- Se esiste un Deployment con selector differente, la reconcile fallisce per evitare adozioni errate.
+
+## Test
+- Eseguito `go test ./...`.

pkg/apis/network.harvesterhci.io/v1alpha1/ippool.go

@@ -119,7 +119,7 @@ type IPPoolStatus struct {
 
 	// +optional
 	// +kubebuilder:validation:Optional
-	AgentPodRef *PodReference `json:"agentPodRef,omitempty"`
+	AgentDeploymentRef *DeploymentReference `json:"agentDeploymentRef,omitempty"`
 
 	// +optional
 	// +kubebuilder:validation:Optional
@@ -132,7 +132,7 @@ type IPv4Status struct {
 	Available int               `json:"available"`
 }
 
-type PodReference struct {
+type DeploymentReference struct {
 	Namespace string    `json:"namespace,omitempty"`
 	Name      string    `json:"name,omitempty"`
 	Image     string    `json:"image,omitempty"`

pkg/apis/network.harvesterhci.io/v1alpha1/zz_generated_deepcopy.go

@@ -9,6 +9,7 @@ import (
 	controllergen "github.com/rancher/wrangler/v3/pkg/controller-gen"
 	"github.com/rancher/wrangler/v3/pkg/controller-gen/args"
 	"github.com/sirupsen/logrus"
+	appsv1 "k8s.io/api/apps/v1"
 	corev1 "k8s.io/api/core/v1"
 	kubevirtv1 "kubevirt.io/api/core/v1"
 )
@@ -48,6 +49,11 @@ func main() {
 					corev1.Pod{},
 				},
 			},
+			appsv1.SchemeGroupVersion.Group: {
+				Types: []interface{}{
+					appsv1.Deployment{},
+				},
+			},
 			cniv1.SchemeGroupVersion.Group: {
 				Types: []interface{}{
 					cniv1.NetworkAttachmentDefinition{},

pkg/codegen/main.go

@@ -24,6 +24,7 @@ import (
 	"github.com/harvester/vm-dhcp-controller/pkg/cache"
 	"github.com/harvester/vm-dhcp-controller/pkg/crd"
 	"github.com/harvester/vm-dhcp-controller/pkg/dhcp"
+	ctlapps "github.com/harvester/vm-dhcp-controller/pkg/generated/controllers/apps"
 	ctlcore "github.com/harvester/vm-dhcp-controller/pkg/generated/controllers/core"
 	ctlcni "github.com/harvester/vm-dhcp-controller/pkg/generated/controllers/k8s.cni.cncf.io"
 	ctlkubevirt "github.com/harvester/vm-dhcp-controller/pkg/generated/controllers/kubevirt.io"
@@ -96,6 +97,7 @@ type Management struct {
 
 	HarvesterNetworkFactory *ctlnetwork.Factory
 
+	AppsFactory     *ctlapps.Factory
 	CniFactory      *ctlcni.Factory
 	CoreFactory     *ctlcore.Factory
 	KubeVirtFactory *ctlkubevirt.Factory
@@ -169,6 +171,13 @@ func SetupManagement(ctx context.Context, restConfig *rest.Config, options *Cont
 	management.CoreFactory = core
 	management.starters = append(management.starters, core)
 
+	apps, err := ctlapps.NewFactoryFromConfigWithOptions(restConfig, opts)
+	if err != nil {
+		return nil, err
+	}
+	management.AppsFactory = apps
+	management.starters = append(management.starters, apps)
+
 	cni, err := ctlcni.NewFactoryFromConfigWithOptions(restConfig, opts)
 	if err != nil {
 		return nil, err