cart and checkout

2026-06-27 19:49:00 +02:00
parent 492f54ff45
commit 528c59bfd3
67 changed files with 3618 additions and 1031 deletions
@@ -0,0 +1,68 @@
+package main
+
+import (
+	"log"
+
+	"git.k6n.net/mats/go-cart-actor/pkg/discovery"
+	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/client-go/kubernetes"
+	"k8s.io/client-go/rest"
+)
+
+// GetDiscovery returns the k8s pod-watcher that finds peer profile replicas, or
+// nil when POD_IP is unset (single-node / local dev — clustering disabled). The
+// watch is scoped to this pod's own namespace so it only needs a namespaced Role
+// (pods: get/list/watch), not a cluster-wide role.
+func GetDiscovery() discovery.Discovery {
+	if podIp == "" {
+		return nil
+	}
+
+	config, kerr := rest.InClusterConfig()
+	if kerr != nil {
+		log.Fatalf("Error creating kubernetes client: %v\n", kerr)
+	}
+	client, err := kubernetes.NewForConfig(config)
+	if err != nil {
+		log.Fatalf("Error creating client: %v\n", err)
+	}
+	timeout := int64(30)
+	return discovery.NewK8sDiscoveryInNamespace(client, discovery.InClusterNamespace(), v1.ListOptions{
+		LabelSelector:  "actor-pool=profile",
+		TimeoutSeconds: &timeout,
+	})
+}
+
+// UseDiscovery starts the watch loop that adds/removes peer hosts from the pool
+// as profile replicas become ready or go away. A nil discovery (POD_IP unset)
+// logs and returns, leaving the pool in single-node mode.
+func UseDiscovery(pool discovery.DiscoveryTarget) {
+	go func(hw discovery.Discovery) {
+		if hw == nil {
+			log.Print("No discovery service available")
+			return
+		}
+		ch, err := hw.Watch()
+		if err != nil {
+			log.Printf("Discovery error: %v", err)
+			return
+		}
+		for evt := range ch {
+			if evt.Host == "" {
+				continue
+			}
+			switch evt.IsReady {
+			case false:
+				if pool.IsKnown(evt.Host) {
+					log.Printf("Host %s is not ready, removing", evt.Host)
+					pool.RemoveHost(evt.Host)
+				}
+			default:
+				if !pool.IsKnown(evt.Host) {
+					log.Printf("Discovered host %s", evt.Host)
+					pool.AddRemoteHost(evt.Host)
+				}
+			}
+		}
+	}(GetDiscovery())
+}
@@ -17,9 +17,39 @@ import (
 	"git.k6n.net/mats/go-cart-actor/internal/ucp"
 	"git.k6n.net/mats/go-cart-actor/pkg/actor"
 	"git.k6n.net/mats/go-cart-actor/pkg/profile"
+	"git.k6n.net/mats/go-cart-actor/pkg/proxy"
+	"git.k6n.net/mats/go-cart-actor/pkg/telemetry"
+	"github.com/redis/go-redis/v9"
 	"go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp"
 )

+// buildAuthStorage selects the credential store and login limiter that back the
+// customer-auth surface. With REDIS_ADDRESS set it uses shared Redis storage so
+// the service scales horizontally; otherwise it falls back to the file-backed
+// store and an in-memory limiter (single-instance / local dev). A Redis failure
+// at startup is fatal rather than silently degrading to per-replica state.
+func buildAuthStorage(ctx context.Context, profileDir string) (customerauth.Credentials, customerauth.Limiter) {
+	if addr := os.Getenv("REDIS_ADDRESS"); addr != "" {
+		rdb := redis.NewClient(&redis.Options{
+			Addr:     addr,
+			Password: os.Getenv("REDIS_PASSWORD"),
+			DB:       0,
+		})
+		store, err := customerauth.NewRedisCredentialStore(ctx, rdb)
+		if err != nil {
+			log.Fatalf("Error connecting customer-auth credential store to Redis at %s: %v", addr, err)
+		}
+		log.Printf("customer-auth: using shared Redis storage at %s", addr)
+		return store, customerauth.NewRedisLoginLimiter(rdb, 0, 0)
+	}
+	credStore, err := customerauth.LoadCredentialStore(filepath.Join(profileDir, "credentials.json"))
+	if err != nil {
+		log.Fatalf("Error loading credential store: %v", err)
+	}
+	log.Print("customer-auth: REDIS_ADDRESS unset — using file credential store + in-memory limiter (single-instance only)")
+	return credStore, customerauth.NewLoginLimiter(0, 0)
+}
+
 // authSecret returns the HMAC key for customer session cookies. It reads
 // CUSTOMER_AUTH_SECRET; when unset it generates an ephemeral random key and
 // warns that issued sessions will not survive a restart (fine for dev).
@@ -86,6 +116,13 @@ func main() {
 		// Destroy is an optional grain-eviction cleanup hook; profiles need none
 		// (disk storage persists via its own loop), so it's left unset — purge()
 		// skips a nil Destroy.
+		// SpawnHost makes the pool cluster-aware: when a grain is owned by a peer
+		// replica, the pool proxies to it over gRPC :1337 (control) + HTTP :8080
+		// (requests). With POD_IP unset and no discovered peers this is never
+		// invoked, so the service still runs fine single-node.
+		SpawnHost: func(host string) (actor.Host[profile.ProfileGrain], error) {
+			return proxy.NewRemoteHost[profile.ProfileGrain](host)
+		},
 		TTL:      10 * time.Minute,
 		PoolSize: 65535,
 		Hostname: podIp,
@@ -96,19 +133,40 @@ func main() {
 		log.Fatalf("Error creating profile pool: %v\n", err)
 	}

+	// Clustering control plane: gRPC server on :1337 serves peer pools'
+	// ownership announcements and remote Apply/Get calls; UseDiscovery watches
+	// for sibling profile pods (label actor-pool=profile) and wires them into the
+	// pool. Both are no-ops in single-node mode (POD_IP unset → nil discovery).
+	grpcSrv, err := actor.NewControlServer[profile.ProfileGrain](actor.DefaultServerConfig(), pool)
+	if err != nil {
+		log.Fatalf("Error starting control plane gRPC server: %v\n", err)
+	}
+	defer grpcSrv.GracefulStop()
+
+	UseDiscovery(pool)
+
 	mux := http.NewServeMux()
 	debugMux := http.NewServeMux()

 	ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt)
 	defer stop()

-	otelShutdown, err := setupOTelSDK(ctx)
+	otelShutdown, err := telemetry.SetupOTelSDK(ctx)
 	if err != nil {
 		log.Fatalf("Unable to start otel %v", err)
 	}

+	// Customer auth storage and failed-login limiter.
+	// Credentials and the failed-login limiter use shared Redis storage when
+	// REDIS_ADDRESS is set (so the service scales horizontally), and fall back to
+	// the file store + in-memory limiter for single-instance / local dev. Session
+	// and verify/reset tokens are stateless HMAC values, so a shared
+	// CUSTOMER_AUTH_SECRET is all they need to work across replicas.
+	credStore, limiter := buildAuthStorage(ctx, profileDir)
+
 	// UCP Customer REST adapter
-	customerUCP := ucp.CustomerHandler(pool)
+	auditLogPath := filepath.Join(profileDir, "audit.log")
+	customerUCP := ucp.CustomerHandler(pool, auditLogPath, credStore)
 	if signer := loadUCPProfileSigner(); signer != nil {
 		customerUCP = ucp.WithSigning(customerUCP, signer)
 		log.Print("ucp customer signing enabled")
@@ -119,11 +177,11 @@ func main() {
 	mux.Handle("/ucp/v1/customers/", http.StripPrefix("/ucp/v1/customers", customerUCP))

 	// Customer auth: password signup/login + session cookies + identity linking.
-	credStore, err := customerauth.LoadCredentialStore(filepath.Join(profileDir, "credentials.json"))
-	if err != nil {
-		log.Fatalf("Error loading credential store: %v\n", err)
-	}
-	authHandler := customerauth.NewServer(credStore, pool, customerauth.NewSigner(authSecret()), 0).Handler()
+	authHandler := customerauth.NewServer(credStore, pool, customerauth.NewSigner(authSecret()), 0, customerauth.Options{
+		Limiter:              limiter,
+		BaseURL:              os.Getenv("CUSTOMER_AUTH_BASE_URL"),
+		RequireVerifiedEmail: os.Getenv("CUSTOMER_AUTH_REQUIRE_VERIFIED") == "true",
+	}).Handler()
 	mux.Handle("/ucp/v1/auth/", http.StripPrefix("/ucp/v1/auth", authHandler))

 	mux.HandleFunc("/healthz", func(w http.ResponseWriter, r *http.Request) {
@@ -1,108 +0,0 @@
-package main
-
-import (
-	"context"
-	"errors"
-	"time"
-
-	"go.opentelemetry.io/otel"
-	"go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploggrpc"
-	"go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc"
-	"go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc"
-	"go.opentelemetry.io/otel/log/global"
-	"go.opentelemetry.io/otel/propagation"
-	"go.opentelemetry.io/otel/sdk/log"
-	"go.opentelemetry.io/otel/sdk/metric"
-	"go.opentelemetry.io/otel/sdk/trace"
-)
-
-// setupOTelSDK bootstraps the OpenTelemetry pipeline.
-// If it does not return an error, make sure to call shutdown for proper cleanup.
-func setupOTelSDK(ctx context.Context) (func(context.Context) error, error) {
-	var shutdownFuncs []func(context.Context) error
-	var err error
-
-	shutdown := func(ctx context.Context) error {
-		var err error
-		for _, fn := range shutdownFuncs {
-			err = errors.Join(err, fn(ctx))
-		}
-		shutdownFuncs = nil
-		return err
-	}
-
-	handleErr := func(inErr error) {
-		err = errors.Join(inErr, shutdown(ctx))
-	}
-
-	prop := newPropagator()
-	otel.SetTextMapPropagator(prop)
-
-	tracerProvider, err := newTracerProvider()
-	if err != nil {
-		handleErr(err)
-		return shutdown, err
-	}
-	shutdownFuncs = append(shutdownFuncs, tracerProvider.Shutdown)
-	otel.SetTracerProvider(tracerProvider)
-
-	meterProvider, err := newMeterProvider()
-	if err != nil {
-		handleErr(err)
-		return shutdown, err
-	}
-	shutdownFuncs = append(shutdownFuncs, meterProvider.Shutdown)
-	otel.SetMeterProvider(meterProvider)
-
-	loggerProvider, err := newLoggerProvider()
-	if err != nil {
-		handleErr(err)
-		return shutdown, err
-	}
-	shutdownFuncs = append(shutdownFuncs, loggerProvider.Shutdown)
-	global.SetLoggerProvider(loggerProvider)
-
-	return shutdown, err
-}
-
-func newPropagator() propagation.TextMapPropagator {
-	return propagation.NewCompositeTextMapPropagator(
-		propagation.TraceContext{},
-		propagation.Baggage{},
-	)
-}
-
-func newTracerProvider() (*trace.TracerProvider, error) {
-	traceExporter, err := otlptracegrpc.New(context.Background())
-	if err != nil {
-		return nil, err
-	}
-
-	tracerProvider := trace.NewTracerProvider(
-		trace.WithBatcher(traceExporter,
-			trace.WithBatchTimeout(time.Second)),
-	)
-	return tracerProvider, nil
-}
-
-func newMeterProvider() (*metric.MeterProvider, error) {
-	exporter, err := otlpmetricgrpc.New(context.Background())
-	if err != nil {
-		return nil, err
-	}
-
-	provider := metric.NewMeterProvider(metric.WithReader(metric.NewPeriodicReader(exporter)))
-	return provider, nil
-}
-
-func newLoggerProvider() (*log.LoggerProvider, error) {
-	logExporter, err := otlploggrpc.New(context.Background())
-	if err != nil {
-		return nil, err
-	}
-
-	loggerProvider := log.NewLoggerProvider(
-		log.WithProcessor(log.NewBatchProcessor(logExporter)),
-	)
-	return loggerProvider, nil
-}