Add new validations to webhooks that reconciliation would prevent

CHANGELOG.md

@@ -11,6 +11,9 @@ Changelog for Cass Operator, new PRs should update the `main / unreleased` secti
 
 ## unreleased
 
+* [ENHANCEMENT] [#912](https://github.com/k8ssandra/cass-operator/issues/912) Add new webhook validations for maxUnavailable string format as well as PVC sizes
+* [ENHANCEMENT] [#902](https://github.com/k8ssandra/cass-operator/issues/902) If scaling down or scaling up process is still ongoing, the webhook will prevent changing the cluster size.
+
 ## v1.30.0
 
 * [CHANGE] [#905](https://github.com/k8ssandra/cass-operator/issues/905) Relax the ServerVersion checks to structure only without separating OSS/DSE/HCD.

apis/cassandra/v1beta1/cassandradatacenter_types.go

@@ -88,6 +88,9 @@ const (
 	// EnableParallelCleanupWithinRackAnnotation speeds up post-scale-out cleanup by processing nodes in parallel within a rack.
 	EnableParallelCleanupWithinRackAnnotation = "cassandra.datastax.com/enable-parallel-cleanup-within-rack"
 
+	// BypassWebhookValidationsAnnotation allows bypassing CassandraDatacenter update webhook validations
+	BypassWebhookValidationsAnnotation = "cassandra.datastax.com/bypass-webhook-validations"
+
 	AllowUpdateAlways AllowUpdateType = "always"
 	AllowUpdateOnce   AllowUpdateType = "once"
 

internal/webhooks/cassandra/v1beta1/cassandradatacenter_webhook.go

@@ -12,13 +12,15 @@ import (
 	apiequality "k8s.io/apimachinery/pkg/api/equality"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/util/intstr"
 	ctrl "sigs.k8s.io/controller-runtime"
 	logf "sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/webhook"
 	"sigs.k8s.io/controller-runtime/pkg/webhook/admission"
 
 	"github.com/google/go-cmp/cmp"
 	api "github.com/k8ssandra/cass-operator/apis/cassandra/v1beta1"
+	corev1 "k8s.io/api/core/v1"
 )
 
 // log is for logging in this package.
@@ -85,6 +87,13 @@ func (v *CassandraDatacenterCustomValidator) ValidateUpdate(ctx context.Context,
 
 	log.Info("Validation for CassandraDatacenter upon update", "name", dc.GetName())
 
+	if metav1.HasAnnotation(dc.ObjectMeta, api.BypassWebhookValidationsAnnotation) &&
+		dc.Annotations[api.BypassWebhookValidationsAnnotation] == "true" {
+		log.Info("Webhook validations bypassed with annotation")
+
+		return nil, nil
+	}
+
 	if err := ValidateSingleDatacenter(dc); err != nil {
 		return nil, err
 	}
@@ -134,6 +143,13 @@ func ValidateSingleDatacenter(dc *api.CassandraDatacenter) error {
 		return attemptedTo("define config dse-yaml with %s", serverStr)
 	}
 
+	if dc.Spec.MaxUnavailable != nil {
+		maxUnavailable, err := intstr.GetScaledValueFromIntOrPercent(dc.Spec.MaxUnavailable, int(dc.Spec.Size), true)
+		if err != nil || maxUnavailable < 0 {
+			return attemptedTo("use invalid maxUnavailable value '%s'", dc.Spec.MaxUnavailable.String())
+		}
+	}
+
 	// if using multiple nodes per worker, requests and limits should be set for both cpu and memory
 	if dc.Spec.AllowMultipleNodesPerWorker {
 		if dc.Spec.Resources.Requests.Cpu().IsZero() ||
@@ -180,6 +196,15 @@ func ValidateDatacenterFieldChanges(oldDc *api.CassandraDatacenter, newDc *api.C
 
 	oldClaimSpec := oldDc.Spec.StorageConfig.CassandraDataVolumeClaimSpec.DeepCopy()
 	newClaimSpec := newDc.Spec.StorageConfig.CassandraDataVolumeClaimSpec.DeepCopy()
+	if oldClaimSpec != nil && newClaimSpec != nil {
+		oldStorageRequest := oldClaimSpec.Resources.Requests[corev1.ResourceStorage]
+		newStorageRequest := newClaimSpec.Resources.Requests[corev1.ResourceStorage]
+
+		if oldStorageRequest.Cmp(newStorageRequest) > 0 {
+			return attemptedTo(
+				"shrink storageConfig.CassandraDataVolumeClaimSpec from %s to %s", oldStorageRequest.String(), newStorageRequest.String())
+		}
+	}
 
 	// CassandraDataVolumeClaimSpec changes are disallowed
 	if metav1.HasAnnotation(newDc.ObjectMeta, api.AllowStorageChangesAnnotation) && newDc.Annotations[api.AllowStorageChangesAnnotation] == "true" {
@@ -192,6 +217,17 @@ func ValidateDatacenterFieldChanges(oldDc *api.CassandraDatacenter, newDc *api.C
 		return attemptedTo("change storageConfig.CassandraDataVolumeClaimSpec, diff: %s", pvcSourceDiff)
 	}
 
+	if oldDc.Spec.Size != newDc.Spec.Size {
+		if oldDc.GetConditionStatus(api.DatacenterScalingUp) == corev1.ConditionTrue {
+			return attemptedTo("change size while datacenter is still scaling up")
+		}
+
+		if oldDc.GetConditionStatus(api.DatacenterScalingDown) == corev1.ConditionTrue {
+			// Here we don't want to allow any changes as scaling down is a lot heavier operation
+			return attemptedTo("change size while datacenter is still scaling down")
+		}
+	}
+
 	// Topology changes - Racks
 	// - Rack Name and Zone changes are disallowed.
 	// - Removing racks is not supported.