terdut-operator/internal/controller/terdutserver_database.go

package controller

import (
	"context"
	"fmt"

	corev1 "k8s.io/api/core/v1"
	apierrors "k8s.io/apimachinery/pkg/api/errors"
	"k8s.io/apimachinery/pkg/api/meta"
	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
	"sigs.k8s.io/controller-runtime/pkg/client"

	terdutv1alpha1 "git.ryuvia.com/niklas/terdut-operator/api/v1alpha1"
)

// databaseError carries a condition reason/message out of resolveDatabaseEnv
// — distinct from a plain error, since "the database isn't ready yet" is an
// expected, requeue-and-retry outcome (DESIGN.md §8), not a reconcile
// failure.
type databaseError struct {
	reason  string
	message string
}

func (e *databaseError) Error() string { return e.message }

// classifyClusterGetError turns a failed Get of the postgresql.acid.zalan.do
// CR into the right condition — distinct from a plain reconcile error, and
// its own function (not inlined) so it's unit-testable against a
// hand-constructed error without needing a real client at all: a
// *meta.NoKindMatchError means the CRD itself isn't installed
// (ReasonPostgresOperatorCRDNotInstalled, §8/§9's graceful-degradation
// case), apierrors.IsNotFound means the CRD exists but this particular
// object doesn't (yet) (ReasonPostgresClusterNotFound).
func classifyClusterGetError(err error, clusterName, namespace string) *databaseError {
	switch {
	case meta.IsNoMatchError(err):
		return &databaseError{
			reason:  terdutv1alpha1.ReasonPostgresOperatorCRDNotInstalled,
			message: "spec.database.postgresClusterRef is set but the postgresql.acid.zalan.do CRD isn't installed in this cluster",
		}
	case apierrors.IsNotFound(err):
		return &databaseError{
			reason:  terdutv1alpha1.ReasonPostgresClusterNotFound,
			message: fmt.Sprintf("postgresql.acid.zalan.do %q not found in namespace %q", clusterName, namespace),
		}
	default:
		// Any other error (RBAC, transient API failure) is a real
		// reconcile error, not a condition to report and wait out --
		// callers still surface it as a databaseError so resolveDatabaseEnv
		// has one return shape, but Reconcile treats every databaseError as
		// a wait-and-retry outcome today. Revisit if that ever matters in
		// practice (a persistent RBAC misconfiguration looks identical to
		// "not ready yet" until someone reads the condition message).
		return &databaseError{reason: terdutv1alpha1.ReasonPostgresClusterNotFound, message: err.Error()}
	}
}

// resolveDatabaseEnv implements DESIGN.md §8's two Postgres paths. It never
// reads a password's value — only ever wires a secretKeyRef into the
// Deployment's env, the same way the chart does — so it needs no RBAC on
// Secret *data* for this, only on the postgresql.acid.zalan.do CR itself
// (the XValidation rule on TerdutServerSpec.Database guarantees exactly one
// of the two fields below is set, so there's no third case to handle).
func (r *TerdutServerReconciler) resolveDatabaseEnv(ctx context.Context, srv *terdutv1alpha1.TerdutServer) ([]corev1.EnvVar, *databaseError) {
	db := srv.Spec.Database

	if db.PostgresClusterRef != nil {
		return r.resolveZalandoDatabaseEnv(ctx, srv, db.PostgresClusterRef.Name)
	}

	env := []corev1.EnvVar{{Name: "TERDUT_DB_DSN", Value: db.DSN}}
	if db.PasswordSecretRef != nil {
		env = append(env, corev1.EnvVar{
			Name: "PGPASSWORD",
			ValueFrom: &corev1.EnvVarSource{
				SecretKeyRef: &corev1.SecretKeySelector{
					LocalObjectReference: corev1.LocalObjectReference{Name: db.PasswordSecretRef.Name},
					Key:                  db.PasswordSecretRef.Key,
				},
			},
		})
	}
	return env, nil
}

// resolveZalandoDatabaseEnv resolves a Zalando postgres-operator
// `postgresql` CR into a DSN + PGPASSWORD secretKeyRef, without reading the
// generated credentials Secret's value — only confirming it exists, the
// same "exists, don't read" posture the DSN path takes toward its own
// password Secret.
//
// By convention (DESIGN.md §8, PostgresClusterRef's own doc comment) the
// database and role are both named "terdut", and the primary Service is
// named after the cluster CR itself — Zalando's own naming convention, not
// something its status exposes as a field to read.
func (r *TerdutServerReconciler) resolveZalandoDatabaseEnv(
	ctx context.Context, srv *terdutv1alpha1.TerdutServer, clusterName string,
) ([]corev1.EnvVar, *databaseError) {
	cluster := &unstructured.Unstructured{}
	cluster.SetGroupVersionKind(postgresqlGVK)
	if err := r.Get(ctx, client.ObjectKey{Namespace: srv.Namespace, Name: clusterName}, cluster); err != nil {
		return nil, classifyClusterGetError(err, clusterName, srv.Namespace)
	}

	credsSecretName := fmt.Sprintf("terdut.%s.credentials.postgresql.acid.zalan.do", clusterName)
	var secret corev1.Secret
	if err := r.Get(ctx, client.ObjectKey{Namespace: srv.Namespace, Name: credsSecretName}, &secret); err != nil {
		if apierrors.IsNotFound(err) {
			return nil, &databaseError{
				reason: terdutv1alpha1.ReasonPostgresClusterNotFound,
				message: fmt.Sprintf(
					"postgresql.acid.zalan.do %q found, but its generated credentials Secret %q doesn't exist yet",
					clusterName, credsSecretName),
			}
		}
		return nil, &databaseError{reason: terdutv1alpha1.ReasonPostgresClusterNotFound, message: err.Error()}
	}

	host := fmt.Sprintf("%s.%s.svc", clusterName, srv.Namespace)
	dsn := fmt.Sprintf("postgres://terdut@%s:5432/terdut?sslmode=require", host)
	return []corev1.EnvVar{
		{Name: "TERDUT_DB_DSN", Value: dsn},
		{
			Name: "PGPASSWORD",
			ValueFrom: &corev1.EnvVarSource{
				SecretKeyRef: &corev1.SecretKeySelector{
					LocalObjectReference: corev1.LocalObjectReference{Name: credsSecretName},
					Key:                  "password",
				},
			},
		},
	}, nil
}