Unify the Tenant CRD with enterprise-api -provision-tenant (lightweight)

Closes a gap named across CLAUDE.md/docs/architecture.md/deploy/README.md
since early Phase 4: the operator's Tenant CRD and -provision-tenant
were two disconnected mechanisms. The operator's reconciler generated a
K8s Secret with a locally-generated random password that authenticated
against nothing (nothing ever called ClickHouse to create a matching
user), and unconditionally claimed status.phase=Active the moment a
Tenant object existed -- actively misleading, not just incomplete.

Two unification shapes were considered (surfaced to the user via
AskUserQuestion, given the real difference in blast radius): the
operator's reconcile loop becoming a second real actor (new Postgres +
ClickHouse admin credentials flowing into the K8s controller, plus real
reconcile-loop idempotency/retry design for an inherently one-shot
external side effect), or keeping -provision-tenant as the sole real
actor and having it also sync its result into the CRD. Went with the
lighter option.

enterprise/internal/tenantcrd (new): a Syncer using the K8s dynamic
client (unstructured.Unstructured + a GroupVersionResource, not
deploy/operator's typed Tenant struct -- avoids a cross-module Go
dependency between two independently-versioned modules for one type).
Upserts the Tenant object, creates/updates a Secret with the *real*
ClickHouse credentials owned by that Tenant via an OwnerReference, then
patches status.{clickHouseDatabaseName,clickHouseSecretRef,
tantivyIndexPath}. Idempotent and safe to retry: never rotates a
credential across a re-sync, never overwrites a pre-existing
spec.displayName a human/GitOps process set.

cmd/enterprise-api/main.go's runProvisionTenant calls Sync when
TENANT_CRD_NAMESPACE is set (empty = no-op, same shape as every other
optional dependency in this codebase). Its "already active" refusal is
now split: ClickHouse re-provisioning is still refused (rotating a live
credential would break every open connection for no benefit), but CR
sync alone is now retryable using the credentials already on file in
rbacstore -- needed for retrying a previously-failed sync, or
backfilling CR sync for a tenant provisioned before this existed.

deploy/operator's reconciler rewritten to match: it never claims
PhaseActive on its own initiative anymore, only once
status.ClickHouseDatabaseName is non-empty (the field -provision-tenant,
and only -provision-tenant, sets). Phase is now a pure function of
{spec.suspended, status.ClickHouseDatabaseName != ""} recomputed every
reconcile, not toggled in place -- fixes a related bug the old code
would have hit once suspension was involved: un-suspending an
already-provisioned tenant needs to return straight to Active, which
isn't derivable from "last observed phase was Suspended" alone. The
reconciler no longer creates or manages any Secret, dropped its
`secrets` RBAC grant entirely, and gained zero new dependencies.

Helm chart: enterprise-api gets its own ServiceAccount/Role/RoleBinding
(get/list/create tenants, get/update/patch tenants/status, get/create/
update secrets -- least-privilege, scoped to the release namespace, not
a ClusterRole) and a TENANT_CRD_NAMESPACE env var, both gated on
tenantOperator.enabled. tenant-operator's ClusterRole loses the
secrets grant it no longer needs.

Verified in this environment: enterprise/internal/tenantcrd's tests run
against k8s.io/client-go's fake dynamic + typed clientsets (real client
library, fake transport, no cluster needed); deploy/operator's rewritten
tenant_controller_test.go runs against controller-runtime's fake
client, including new regression tests for the "must not claim Active
without confirmation" and "un-suspending returns to Active, not
Provisioning" properties; helm template + parsing the rendered YAML
confirms the RBAC split renders exactly as designed under both
tenantOperator.enabled=true/false. Not verified: an actual
-provision-tenant run against a real cluster with the operator watching
(no live cluster in this environment, same disclosed limitation as the
rest of /deploy). Docs updated in lockstep: CLAUDE.md, docs/architecture.md,
deploy/README.md, deploy/helm/sentry/README.md (including a corrected
"Trying the two-tenant example" walkthrough), phase-4-runbook.md (new
§11), enterprise/README.md. Also fixed two unrelated stale claims found
along the way: docs/architecture.md still said docker-compose.yml ran
plain api unconditionally (fixed in an earlier commit, doc not updated
then), and enterprise-api's own main.go doc comment still said Helm/
docker-compose wiring wasn't built yet.
This commit is contained in:
2026-08-14 09:07:10 -07:00
parent 8d7326fc6a
commit 823f5d48d1
18 changed files with 1073 additions and 323 deletions
+229
View File
@@ -0,0 +1,229 @@
// Package tenantcrd syncs enterprise-api -provision-tenant's real
// provisioning result into the Tenant CRD (deploy/operator/api/
// v1alpha1) -- the "lightweight unification" named in CLAUDE.md's "two
// independent provisioning mechanisms" gap: -provision-tenant stays the
// sole real actor (rbacstore + tenantprovision.ProvisionClickHouse, see
// runProvisionTenant's doc comment in cmd/enterprise-api/main.go); this
// package's only job is making that result observable via `kubectl get
// tenants` too, for a deployment that also runs deploy/operator's
// tenant-operator. deploy/operator/internal/controller.TenantReconciler
// derives Phase/Conditions from what this package writes -- it never
// invents "provisioned" on its own.
//
// Deliberately uses the K8s dynamic client (unstructured.Unstructured +
// a GroupVersionResource) rather than importing deploy/operator/api/
// v1alpha1's typed Tenant struct: deploy/operator is a separate Go
// module (its own go.mod), and adding a cross-module `replace` directive
// between two independently-versioned modules is exactly the kind of
// coupling this codebase has otherwise avoided (enterprise/ already only
// imports api/, never deploy/operator/). The typed Secret API
// (k8s.io/client-go/kubernetes) is used for the credential Secret
// itself, since Secret is a stable, well-known built-in type with no
// such module-boundary concern.
package tenantcrd
import (
"context"
"fmt"
corev1 "k8s.io/api/core/v1"
apierrors "k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
"k8s.io/apimachinery/pkg/runtime/schema"
"k8s.io/apimachinery/pkg/types"
"k8s.io/client-go/dynamic"
"k8s.io/client-go/kubernetes"
"k8s.io/client-go/rest"
"k8s.io/client-go/tools/clientcmd"
)
// tenantGVR identifies deploy/operator/config/crd/sentry.io_tenants.yaml's
// resource -- kept as a plain schema.GroupVersionResource (not the typed
// package) for the reason this file's doc comment explains.
var tenantGVR = schema.GroupVersionResource{Group: "sentry.io", Version: "v1alpha1", Resource: "tenants"}
// Syncer talks to the K8s API. Construction (New) is the only place
// that can fail for "no cluster reachable" reasons -- Sync itself
// assumes a working client.
type Syncer struct {
dynamic dynamic.Interface
clientset kubernetes.Interface
namespace string
}
// New builds a Syncer for namespace, trying in-cluster config first (the
// real deployment shape: -provision-tenant runs via `kubectl exec` into
// the already-running enterprise-api Pod, which carries its own
// ServiceAccount) and falling back to KUBECONFIG/the default kubeconfig
// path for local/dev convenience. Returns an error if neither is
// reachable -- callers decide whether that's fatal (see
// cmd/enterprise-api/main.go's runProvisionTenant: it is, once CR sync
// has been explicitly requested via TENANT_CRD_NAMESPACE, since a
// silent skip would leave the CRD stale/wrong again, the exact bug this
// package exists to fix).
func New(namespace string) (*Syncer, error) {
config, err := rest.InClusterConfig()
if err != nil {
config, err = clientcmd.NewNonInteractiveDeferredLoadingClientConfig(
clientcmd.NewDefaultClientConfigLoadingRules(),
&clientcmd.ConfigOverrides{},
).ClientConfig()
if err != nil {
return nil, fmt.Errorf("tenantcrd: no in-cluster config and no usable kubeconfig: %w", err)
}
}
dyn, err := dynamic.NewForConfig(config)
if err != nil {
return nil, fmt.Errorf("tenantcrd: building dynamic client: %w", err)
}
clientset, err := kubernetes.NewForConfig(config)
if err != nil {
return nil, fmt.Errorf("tenantcrd: building clientset: %w", err)
}
return &Syncer{dynamic: dyn, clientset: clientset, namespace: namespace}, nil
}
// newForTest builds a Syncer directly from fake clients -- production
// code always goes through New, but tests need to inject
// k8s.io/client-go/dynamic/fake and kubernetes/fake without a real or
// in-cluster API server.
func newForTest(dyn dynamic.Interface, clientset kubernetes.Interface, namespace string) *Syncer {
return &Syncer{dynamic: dyn, clientset: clientset, namespace: namespace}
}
// Credentials is the minimal shape Sync needs -- deliberately not
// tenantprovision.Credentials itself, matching chrunner.DataSource's
// "narrow, not the storage type" precedent.
type Credentials struct {
Username string
Password string
}
// SecretName is deterministic from the tenant ID -- never randomly
// suffixed -- so a re-run of Sync (idempotent retry, see
// runProvisionTenant) finds and updates the same Secret rather than
// creating a second one.
func SecretName(tenantID string) string {
return fmt.Sprintf("sentry-tenant-%s-clickhouse", tenantID)
}
// Sync upserts the Tenant object (creating it with spec.displayName if
// absent) and its credential Secret, then patches status to reflect
// real, already-confirmed provisioning -- called only after
// tenantprovision.ProvisionClickHouse has actually succeeded, never
// before. Safe to call repeatedly for the same tenant (idempotent):
// re-running overwrites the Secret with the same credentials rbacstore
// already has on file (never rotates to a *new*, different credential --
// that would break every open connection for no benefit, same
// reasoning the operator's now-removed reconcileSecret documented) and
// re-applies the same status fields.
func (s *Syncer) Sync(ctx context.Context, tenantID, displayName, tantivyIndexPath string, creds Credentials) error {
uid, err := s.upsertTenant(ctx, tenantID, displayName)
if err != nil {
return fmt.Errorf("tenantcrd: upserting tenant object: %w", err)
}
secretName := SecretName(tenantID)
if err := s.upsertSecret(ctx, tenantID, secretName, uid, creds); err != nil {
return fmt.Errorf("tenantcrd: upserting credential secret: %w", err)
}
if err := s.patchStatus(ctx, tenantID, secretName, tantivyIndexPath); err != nil {
return fmt.Errorf("tenantcrd: patching tenant status: %w", err)
}
return nil
}
func (s *Syncer) upsertTenant(ctx context.Context, tenantID, displayName string) (types.UID, error) {
client := s.dynamic.Resource(tenantGVR).Namespace(s.namespace)
existing, err := client.Get(ctx, tenantID, metav1.GetOptions{})
if err == nil {
return existing.GetUID(), nil
}
if !apierrors.IsNotFound(err) {
return "", fmt.Errorf("getting existing tenant object: %w", err)
}
obj := &unstructured.Unstructured{Object: map[string]interface{}{
"apiVersion": "sentry.io/v1alpha1",
"kind": "Tenant",
"metadata": map[string]interface{}{
"name": tenantID,
"namespace": s.namespace,
},
"spec": map[string]interface{}{
"displayName": displayName,
},
}}
created, err := client.Create(ctx, obj, metav1.CreateOptions{})
if err != nil {
return "", fmt.Errorf("creating tenant object: %w", err)
}
return created.GetUID(), nil
}
func (s *Syncer) upsertSecret(ctx context.Context, tenantID, secretName string, tenantUID types.UID, creds Credentials) error {
secrets := s.clientset.CoreV1().Secrets(s.namespace)
controllerTrue := true
secret := &corev1.Secret{
ObjectMeta: metav1.ObjectMeta{
Name: secretName,
Namespace: s.namespace,
Labels: map[string]string{
"app.kubernetes.io/managed-by": "sentry-enterprise-api",
"sentry.io/tenant": tenantID,
},
// Owned by the Tenant object even though a different
// process (this one, not the operator's controller)
// created it -- K8s's garbage collector honors
// OwnerReferences regardless of which actor set them, so
// deleting the Tenant still cleans this Secret up.
OwnerReferences: []metav1.OwnerReference{{
APIVersion: "sentry.io/v1alpha1",
Kind: "Tenant",
Name: tenantID,
UID: tenantUID,
Controller: &controllerTrue,
BlockOwnerDeletion: &controllerTrue,
}},
},
Type: corev1.SecretTypeOpaque,
StringData: map[string]string{
"username": creds.Username,
"password": creds.Password,
"database": tenantID,
},
}
existing, err := secrets.Get(ctx, secretName, metav1.GetOptions{})
switch {
case err == nil:
secret.ResourceVersion = existing.ResourceVersion
_, err := secrets.Update(ctx, secret, metav1.UpdateOptions{})
return err
case apierrors.IsNotFound(err):
_, err := secrets.Create(ctx, secret, metav1.CreateOptions{})
return err
default:
return fmt.Errorf("getting existing secret: %w", err)
}
}
func (s *Syncer) patchStatus(ctx context.Context, tenantID, secretName, tantivyIndexPath string) error {
client := s.dynamic.Resource(tenantGVR).Namespace(s.namespace)
existing, err := client.Get(ctx, tenantID, metav1.GetOptions{})
if err != nil {
return fmt.Errorf("getting tenant object for status update: %w", err)
}
if err := unstructured.SetNestedField(existing.Object, tenantID, "status", "clickHouseDatabaseName"); err != nil {
return err
}
if err := unstructured.SetNestedField(existing.Object, secretName, "status", "clickHouseSecretRef"); err != nil {
return err
}
if err := unstructured.SetNestedField(existing.Object, tantivyIndexPath, "status", "tantivyIndexPath"); err != nil {
return err
}
_, err = client.UpdateStatus(ctx, existing, metav1.UpdateOptions{})
return err
}