Files
fed61d2815 feat: replicating resources upon namespace creation (#2080)
* feat: replicating resources upon namespace creation

This feature enhances the (Global)TenantResource replication by
triggering a replication of resources upon a Namespace creation: this
speeds up the replication of resources without waiting for the
resyncPeriod that could be delayed for several reasons.

Signed-off-by: Dario Tranchitella <dario@tranchitella.eu>

* fix: addressing fix from github copilot

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Signed-off-by: Dario Tranchitella <dario@tranchitella.eu>

---------

Signed-off-by: Dario Tranchitella <dario@tranchitella.eu>
Co-authored-by: Oliver Bähler <26610571+oliverbaehler@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
2026-08-17 10:51:09 +02:00

109 lines
3.5 KiB
Go

// Copyright 2020-2026 Project Capsule Authors
// SPDX-License-Identifier: Apache-2.0
package resources
import (
"context"
"reflect"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/client-go/util/retry"
"sigs.k8s.io/controller-runtime/pkg/client"
capsulev1beta2 "github.com/projectcapsule/capsule/api/v1beta2"
"github.com/projectcapsule/capsule/pkg/api/meta"
"github.com/projectcapsule/capsule/pkg/api/processor"
)
// Grafts the processed items of a single Namespace scope on the status of a replication
// resource, leaving the items of every other Namespace untouched.
//
// The type parameter keeps the accessors bound to the resource they belong to, whereas
// the patching itself is shared: the status holding the processed items is the very same
// for the namespaced and the global replication resources.
type scopedStatusPatcher[T client.Object] struct {
client client.Client
reader client.Reader
// Builds an empty instance to read the latest state into.
factory func() T
// Accesses the status shared by every replication resource.
status func(T) *capsulev1beta2.TenantResourceCommonStatus
}
// Patch merges the given items in the status of the resource,
// reporting them for the given scope only.
// The in memory instance is kept aligned with what has been written.
func (p scopedStatusPatcher[T]) Patch(
ctx context.Context,
instance T,
scope processor.Scope,
items meta.ProcessedItems,
) error {
return retry.RetryOnConflict(retry.DefaultBackoff, func() error {
latest := p.factory()
// The uncached reader is mandatory here: a stale list of processed items would
// resurrect the ones the full reconciliation just pruned.
if err := p.reader.Get(ctx, client.ObjectKeyFromObject(instance), latest); err != nil {
return client.IgnoreNotFound(err)
}
status := p.status(latest)
original := status.DeepCopy()
status.ProcessedItems.ReplaceScope(scope.Tenant, scope.Namespace, items)
status.UpdateStats()
// Only the shared status is compared, since it is the only one being mutated: the
// resource specific part cannot have drifted.
if reflect.DeepEqual(*original, *status) {
return nil
}
if err := p.client.Status().Update(ctx, latest); err != nil {
return err
}
// Keep the in-memory object aligned with what we just wrote.
*p.status(instance) = *status
return nil
})
}
// Builds the processor options out of the settings shared by every replication resource.
//
// The field owner must be derived exactly as the full reconciliation does, otherwise the
// two would fight over the server-side apply ownership of the very same objects.
func scopedProcessorOptions(
obj client.Object,
spec *capsulev1beta2.TenantResourceCommonSpec,
owner *metav1.OwnerReference,
) processor.ProcessorOptions {
return processor.ProcessorOptions{
FieldOwnerPrefix: getFieldOwner(obj.GetName(), obj.GetNamespace()),
Prune: *spec.PruningOnDelete,
Adopt: *spec.Settings.Adopt,
Force: *spec.Settings.Force,
Owner: owner,
}
}
// States whether the given replication resource is out of the Namespace scoped replication
// duties, along with the reason.
// Replicating the same mechanism of reconciling GlobalTenantResource or TenantResource objects.
func skipScopedReplication(
obj client.Object,
spec *capsulev1beta2.TenantResourceCommonSpec,
) (bool, string) {
if !obj.GetDeletionTimestamp().IsZero() {
return true, "is being deleted"
}
if spec.IsCordoned() {
return true, "is cordoned"
}
return false, ""
}