Release core 0.0.40 and bundles to 0.0.53

cmd/main.go

@@ -509,7 +509,6 @@ func main() {
 			Client:  multiclusterClient,
 			Scheme:  mgr.GetScheme(),
 			Monitor: monitor,
-			Conf:    conf.GetConfigOrDie[openstack.OpenStackDatasourceReconcilerConfig](),
 		}).SetupWithManager(mgr, multiclusterClient); err != nil {
 			setupLog.Error(err, "unable to create controller", "controller", "OpenStackDatasourceReconciler")
 			os.Exit(1)

helm/bundles/cortex-cinder/Chart.yaml

@@ -5,7 +5,7 @@ apiVersion: v2
 name: cortex-cinder
 description: A Helm chart deploying Cortex for Cinder.
 type: application
-version: 0.0.52
+version: 0.0.53
 appVersion: 0.1.0
 dependencies:
   # from: file://../../library/cortex-postgres
@@ -16,12 +16,12 @@ dependencies:
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
     alias: cortex-knowledge-controllers
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
     alias: cortex-scheduling-controllers
 
   # Owner info adds a configmap to the kubernetes cluster with information on

helm/bundles/cortex-crds/Chart.yaml

@@ -5,13 +5,13 @@ apiVersion: v2
 name: cortex-crds
 description: A Helm chart deploying Cortex CRDs.
 type: application
-version: 0.0.52
+version: 0.0.53
 appVersion: 0.1.0
 dependencies:
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
 
   # Owner info adds a configmap to the kubernetes cluster with information on
   # the service owner. This makes it easier to find out who to contact in case

helm/bundles/cortex-ironcore/Chart.yaml

@@ -5,13 +5,13 @@ apiVersion: v2
 name: cortex-ironcore
 description: A Helm chart deploying Cortex for IronCore.
 type: application
-version: 0.0.52
+version: 0.0.53
 appVersion: 0.1.0
 dependencies:
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
 
   # Owner info adds a configmap to the kubernetes cluster with information on
   # the service owner. This makes it easier to find out who to contact in case

helm/bundles/cortex-manila/Chart.yaml

@@ -5,7 +5,7 @@ apiVersion: v2
 name: cortex-manila
 description: A Helm chart deploying Cortex for Manila.
 type: application
-version: 0.0.52
+version: 0.0.53
 appVersion: 0.1.0
 dependencies:
   # from: file://../../library/cortex-postgres
@@ -16,12 +16,12 @@ dependencies:
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
     alias: cortex-knowledge-controllers
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
     alias: cortex-scheduling-controllers
 
   # Owner info adds a configmap to the kubernetes cluster with information on

helm/bundles/cortex-nova/Chart.yaml

@@ -5,7 +5,7 @@ apiVersion: v2
 name: cortex-nova
 description: A Helm chart deploying Cortex for Nova.
 type: application
-version: 0.0.52
+version: 0.0.53
 appVersion: 0.1.0
 dependencies:
   # from: file://../../library/cortex-postgres
@@ -16,12 +16,12 @@ dependencies:
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
     alias: cortex-knowledge-controllers
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
     alias: cortex-scheduling-controllers
 
   # Owner info adds a configmap to the kubernetes cluster with information on

helm/bundles/cortex-nova/alerts/nova.alerts.yaml

@@ -627,7 +627,9 @@ groups:
         the datasource controller's workqueue overprioritizing other datasources.
 
   - alert: CortexNovaExistingDatasourcesLackingBehind
-    expr: sum by(datasource) (cortex_datasource_seconds_until_reconcile{queued="true",domain="nova"}) < -600
+    expr: |
+      sum by(datasource) (cortex_datasource_seconds_until_reconcile{queued="true",domain="nova"}) < -600
+      and on(datasource) cortex_datasource_state{state="ready",domain="nova"} == 1
     for: 10m
     labels:
       context: datasources

helm/bundles/cortex-pods/Chart.yaml

@@ -5,13 +5,13 @@ apiVersion: v2
 name: cortex-pods
 description: A Helm chart deploying Cortex for Pods.
 type: application
-version: 0.0.52
+version: 0.0.53
 appVersion: 0.1.0
 dependencies:
   # from: file://../../library/cortex
   - name: cortex
     repository: oci://ghcr.io/cobaltcore-dev/cortex/charts
-    version: 0.0.39
+    version: 0.0.40
 
   # Owner info adds a configmap to the kubernetes cluster with information on
   # the service owner. This makes it easier to find out who to contact in case

helm/library/cortex/Chart.yaml

@@ -2,7 +2,7 @@ apiVersion: v2
 name: cortex
 description: A Helm chart to distribute cortex.
 type: application
-version: 0.0.39
-appVersion: "sha-f437366b"
+version: 0.0.40
+appVersion: "sha-be8840bc"
 icon: "https://example.com/icon.png"
 dependencies: []

internal/knowledge/datasources/plugins/openstack/controller.go

@@ -12,6 +12,7 @@ import (
 	"github.com/cobaltcore-dev/cortex/api/v1alpha1"
 	"github.com/cobaltcore-dev/cortex/internal/knowledge/datasources"
 	"github.com/cobaltcore-dev/cortex/internal/knowledge/db"
+	"github.com/cobaltcore-dev/cortex/pkg/conf"
 	"github.com/cobaltcore-dev/cortex/pkg/keystone"
 	"github.com/cobaltcore-dev/cortex/pkg/multicluster"
 	"github.com/cobaltcore-dev/cortex/pkg/sso"
@@ -24,20 +25,25 @@ import (
 	"k8s.io/apimachinery/pkg/runtime"
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/controller"
 	"sigs.k8s.io/controller-runtime/pkg/event"
 	"sigs.k8s.io/controller-runtime/pkg/handler"
 	logf "sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/manager"
 	"sigs.k8s.io/controller-runtime/pkg/predicate"
+	"sigs.k8s.io/controller-runtime/pkg/reconcile"
 )
 
-type OpenStackDatasourceReconcilerConfig struct {
+type config struct {
 	// The controller will only touch resources with this scheduling domain.
 	SchedulingDomain v1alpha1.SchedulingDomain `json:"schedulingDomain"`
 	// Secret ref to keystone credentials stored in a k8s secret.
 	KeystoneSecretRef corev1.SecretReference `json:"keystoneSecretRef"`
 	// Secret ref to SSO credentials stored in a k8s secret, if applicable.
 	SSOSecretRef *corev1.SecretReference `json:"ssoSecretRef"`
+	// The number of parallel reconciles to allow for the controller.
+	// By default, this will be set to 1.
+	ParallelReconciles *int `json:"openstackDatasourceControllerParallelReconciles,omitempty"`
 }
 
 type Syncer interface {
@@ -54,8 +60,9 @@ type OpenStackDatasourceReconciler struct {
 	Scheme *runtime.Scheme
 	// Datasources monitor.
 	Monitor datasources.Monitor
+
 	// Config for the reconciler.
-	Conf OpenStackDatasourceReconcilerConfig
+	conf config
 }
 
 // Reconcile is part of the main kubernetes reconciliation loop which aims to
@@ -281,16 +288,21 @@ func predicateIgnoreStatusConditions() predicate.Predicate {
 }
 
 func (r *OpenStackDatasourceReconciler) SetupWithManager(mgr manager.Manager, mcl *multicluster.Client) error {
+	var err error
+	r.conf, err = conf.GetConfig[config]()
+	if err != nil {
+		return err
+	}
 	bldr := multicluster.BuildController(mcl, mgr)
 	// Watch datasource changes across all clusters.
-	bldr, err := bldr.WatchesMulticluster(
+	bldr, err = bldr.WatchesMulticluster(
 		&v1alpha1.Datasource{},
 		&handler.EnqueueRequestForObject{},
 		predicate.NewPredicateFuncs(func(obj client.Object) bool {
 			// Only react to datasources matching the operator.
 			ds := obj.(*v1alpha1.Datasource)
 			// Ignore all datasources outside our scheduling domain.
-			if ds.Spec.SchedulingDomain != r.Conf.SchedulingDomain {
+			if ds.Spec.SchedulingDomain != r.conf.SchedulingDomain {
 				return false
 			}
 			// Ignore all datasources that are not of type openstack.
@@ -305,5 +317,14 @@ func (r *OpenStackDatasourceReconciler) SetupWithManager(mgr manager.Manager, mc
 		return err
 	}
 	return bldr.Named("cortex-openstack-datasource").
+		WithOptions(controller.TypedOptions[reconcile.Request]{
+			// Allow parallel reconciles if configured, otherwise default to 1.
+			MaxConcurrentReconciles: func() int {
+				if r.conf.ParallelReconciles != nil {
+					return *r.conf.ParallelReconciles
+				}
+				return 1
+			}(),
+		}).
 		Complete(r)
 }

internal/knowledge/datasources/plugins/openstack/controller_test.go

@@ -32,7 +32,7 @@ func TestOpenStackDatasourceReconciler_Creation(t *testing.T) {
 		Client:  client,
 		Scheme:  scheme,
 		Monitor: datasources.Monitor{},
-		Conf:    OpenStackDatasourceReconcilerConfig{SchedulingDomain: "test-operator"},
+		conf:    config{SchedulingDomain: "test-operator"},
 	}
 
 	if reconciler.Client == nil {
@@ -43,8 +43,8 @@ func TestOpenStackDatasourceReconciler_Creation(t *testing.T) {
 		t.Error("Scheme should not be nil")
 	}
 
-	if reconciler.Conf.SchedulingDomain != "test-operator" {
-		t.Errorf("Expected scheduling domain 'test-operator', got %s", reconciler.Conf.SchedulingDomain)
+	if reconciler.conf.SchedulingDomain != "test-operator" {
+		t.Errorf("Expected scheduling domain 'test-operator', got %s", reconciler.conf.SchedulingDomain)
 	}
 }
 

internal/knowledge/datasources/plugins/openstack/nova/nova_api.go

@@ -325,7 +325,7 @@ func (api *novaAPI) GetAllMigrations(ctx context.Context) ([]Migration, error) {
 	initialURL := api.sc.Endpoint + "os-migrations"
 	var nextURL = &initialURL
 	var migrations []Migration
-	seen := make(map[string]struct{})
+	seen := make(map[int]struct{})
 	for nextURL != nil {
 		req, err := http.NewRequestWithContext(ctx, http.MethodGet, *nextURL, http.NoBody)
 		if err != nil {
@@ -354,11 +354,11 @@ func (api *novaAPI) GetAllMigrations(ctx context.Context) ([]Migration, error) {
 			return nil, err
 		}
 		for _, m := range list.Migrations {
-			if _, ok := seen[m.UUID]; ok {
-				slog.Warn("skipping duplicate migration", "uuid", m.UUID)
+			if _, ok := seen[m.ID]; ok {
+				slog.Warn("skipping duplicate migration", "id", m.ID)
 				continue
 			}
-			seen[m.UUID] = struct{}{}
+			seen[m.ID] = struct{}{}
 			migrations = append(migrations, m)
 		}
 		nextURL = nil

internal/scheduling/cinder/external_scheduler_api.go

@@ -93,7 +93,7 @@ func (httpAPI *httpAPI) CinderExternalScheduler(w http.ResponseWriter, r *http.R
 	// Exit early if the request method is not POST.
 	if r.Method != http.MethodPost {
 		internalErr := fmt.Errorf("invalid request method: %s", r.Method)
-		c.Respond(http.StatusMethodNotAllowed, internalErr, "invalid request method")
+		c.Respond(nil, http.StatusMethodNotAllowed, internalErr, "invalid request method")
 		return
 	}
 
@@ -103,7 +103,7 @@ func (httpAPI *httpAPI) CinderExternalScheduler(w http.ResponseWriter, r *http.R
 	// If configured, log out the complete request body.
 	body, err := io.ReadAll(r.Body)
 	if err != nil {
-		c.Respond(http.StatusInternalServerError, err, "failed to read request body")
+		c.Respond(nil, http.StatusInternalServerError, err, "failed to read request body")
 		return
 	}
 	raw := runtime.RawExtension{Raw: body}
@@ -112,17 +112,15 @@ func (httpAPI *httpAPI) CinderExternalScheduler(w http.ResponseWriter, r *http.R
 	cp := body
 	reader := bytes.NewReader(cp)
 	if err := json.NewDecoder(reader).Decode(&requestData); err != nil {
-		c.Respond(http.StatusBadRequest, err, "failed to decode request body")
+		c.Respond(nil, http.StatusBadRequest, err, "failed to decode request body")
 		return
 	}
-	slog.Info(
-		"handling POST request", "url", "/scheduler/cinder/external",
-		"hosts", len(requestData.Hosts), "spec", requestData.Spec,
-	)
+	logger := slog.With(requestData.GetTraceLogArgs())
+	logger.Info("handling POST request", "url", "/scheduler/cinder/external", "body", string(body))
 
 	if ok, reason := httpAPI.canRunScheduler(requestData); !ok {
 		internalErr := fmt.Errorf("cannot run scheduler: %s", reason)
-		c.Respond(http.StatusBadRequest, internalErr, reason)
+		c.Respond(logger, http.StatusBadRequest, internalErr, reason)
 		return
 	}
 
@@ -131,10 +129,10 @@ func (httpAPI *httpAPI) CinderExternalScheduler(w http.ResponseWriter, r *http.R
 		var err error
 		requestData.Pipeline, err = httpAPI.inferPipelineName(requestData)
 		if err != nil {
-			c.Respond(http.StatusBadRequest, err, err.Error())
+			c.Respond(logger, http.StatusBadRequest, err, err.Error())
 			return
 		}
-		slog.Info("inferred pipeline name", "pipeline", requestData.Pipeline)
+		logger.Info("inferred pipeline name", "pipeline", requestData.Pipeline)
 	}
 
 	// Create the decision object in kubernetes.
@@ -154,24 +152,24 @@ func (httpAPI *httpAPI) CinderExternalScheduler(w http.ResponseWriter, r *http.R
 	}
 	ctx := r.Context()
 	if err := httpAPI.delegate.ProcessNewDecisionFromAPI(ctx, decision); err != nil {
-		c.Respond(http.StatusInternalServerError, err, "failed to process scheduling decision")
+		c.Respond(logger, http.StatusInternalServerError, err, "failed to process scheduling decision")
 		return
 	}
 	// Check if the decision contains status conditions indicating an error.
 	if meta.IsStatusConditionFalse(decision.Status.Conditions, v1alpha1.DecisionConditionReady) {
-		c.Respond(http.StatusInternalServerError, errors.New("decision contains error condition"), "decision failed")
+		c.Respond(logger, http.StatusInternalServerError, errors.New("decision contains error condition"), "decision failed")
 		return
 	}
 	if decision.Status.Result == nil {
-		c.Respond(http.StatusInternalServerError, errors.New("decision didn't produce a result"), "decision failed")
+		c.Respond(logger, http.StatusInternalServerError, errors.New("decision didn't produce a result"), "decision failed")
 		return
 	}
 	hosts := decision.Status.Result.OrderedHosts
 	response := api.ExternalSchedulerResponse{Hosts: hosts}
 	w.Header().Set("Content-Type", "application/json")
 	if err = json.NewEncoder(w).Encode(response); err != nil {
-		c.Respond(http.StatusInternalServerError, err, "failed to encode response")
+		c.Respond(logger, http.StatusInternalServerError, err, "failed to encode response")
 		return
 	}
-	c.Respond(http.StatusOK, nil, "Success")
+	c.Respond(logger, http.StatusOK, nil, "Success")
 }

internal/scheduling/lib/api_monitor.go

@@ -53,7 +53,7 @@ func (m *APIMonitor) Callback(w http.ResponseWriter, r *http.Request, pattern st
 
 // Respond to the request with the given code and error.
 // Also log the time it took to handle the request.
-func (c MonitoredCallback) Respond(code int, err error, text string) {
+func (c MonitoredCallback) Respond(logger *slog.Logger, code int, err error, text string) {
 	if c.apiMonitor != nil && c.apiMonitor.ApiRequestsTimer != nil {
 		observer := c.apiMonitor.ApiRequestsTimer.WithLabelValues(
 			c.r.Method,
@@ -64,7 +64,11 @@ func (c MonitoredCallback) Respond(code int, err error, text string) {
 		observer.Observe(time.Since(c.t).Seconds())
 	}
 	if err != nil {
-		slog.Error("failed to handle request", "error", err)
+		if logger == nil {
+			slog.Error("failed to handle request", "error", err)
+		} else {
+			logger.Error("failed to handle request", "error", err)
+		}
 		http.Error(c.w, text, code)
 		return
 	}