Merge branch 'main' into add-atomic-cache-ops

Split the memory and Redis cache implementations into separate files and
provide backend-native atomic set-if-absent support.

management/internals/server/boot.go

@@ -20,17 +20,15 @@ import (
 	"google.golang.org/grpc/credentials"
 	"google.golang.org/grpc/keepalive"
 
-	cachestore "github.com/eko/gocache/lib/v4/store"
-
 	"github.com/netbirdio/netbird/encryption"
 	"github.com/netbirdio/netbird/formatter/hook"
+	"github.com/netbirdio/netbird/management/internals/modules/agentnetwork"
 	"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/accesslogs"
 	accesslogsmanager "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/accesslogs/manager"
 	rpservice "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/service"
 	nbgrpc "github.com/netbirdio/netbird/management/internals/shared/grpc"
 	"github.com/netbirdio/netbird/management/server/activity"
 	activitystore "github.com/netbirdio/netbird/management/server/activity/store"
-	"github.com/netbirdio/netbird/management/internals/modules/agentnetwork"
 	nbcache "github.com/netbirdio/netbird/management/server/cache"
 	nbContext "github.com/netbirdio/netbird/management/server/context"
 	nbhttp "github.com/netbirdio/netbird/management/server/http"
@@ -70,8 +68,8 @@ func (s *BaseServer) Metrics() telemetry.AppMetrics {
 
 // CacheStore returns a shared cache store backed by Redis or in-memory depending on the environment.
 // All consumers should reuse this store to avoid creating multiple Redis connections.
-func (s *BaseServer) CacheStore() cachestore.StoreInterface {
-	return Create(s, func() cachestore.StoreInterface {
+func (s *BaseServer) CacheStore() nbcache.Store {
+	return Create(s, func() nbcache.Store {
 		cs, err := nbcache.NewStore(context.Background(), nbcache.DefaultStoreMaxTimeout, nbcache.DefaultStoreCleanupInterval, nbcache.DefaultStoreMaxConn)
 		if err != nil {
 			log.Fatalf("failed to create shared cache store: %v", err)

management/server/auth/session.go

@@ -7,9 +7,6 @@ import (
 	"errors"
 	"fmt"
 	"time"
-
-	"github.com/eko/gocache/lib/v4/cache"
-	"github.com/eko/gocache/lib/v4/store"
 )
 
 const (
@@ -22,12 +19,17 @@ var (
 	ErrTokenExpired     = errors.New("JWT expired")
 )
 
-type SessionStore struct {
-	cache *cache.Cache[string]
+// TokenCache atomically records used JWTs until their expiration.
+type TokenCache interface {
+	SetNX(ctx context.Context, key, value string, ttl time.Duration) (bool, error)
 }
 
-func NewSessionStore(cacheStore store.StoreInterface) *SessionStore {
-	return &SessionStore{cache: cache.New[string](cacheStore)}
+type SessionStore struct {
+	cache TokenCache
+}
+
+func NewSessionStore(cacheStore TokenCache) *SessionStore {
+	return &SessionStore{cache: cacheStore}
 }
 
 // RegisterToken records a JWT until its exp time and rejects reuse.
@@ -38,19 +40,13 @@ func (s *SessionStore) RegisterToken(ctx context.Context, token string, expiresA
 	}
 
 	key := usedTokenKeyPrefix + hashToken(token)
-	_, err := s.cache.Get(ctx, key)
-	if err == nil {
-		return ErrTokenAlreadyUsed
-	}
-
-	var notFound *store.NotFound
-	if !errors.As(err, &notFound) {
-		return fmt.Errorf("failed to lookup used token entry: %w", err)
-	}
-
-	if err := s.cache.Set(ctx, key, usedTokenMarker, store.WithExpiration(ttl)); err != nil {
+	created, err := s.cache.SetNX(ctx, key, usedTokenMarker, ttl)
+	if err != nil {
 		return fmt.Errorf("failed to store used token entry: %w", err)
 	}
+	if !created {
+		return ErrTokenAlreadyUsed
+	}
 
 	return nil
 }

management/server/auth/session_test.go

@@ -2,6 +2,7 @@ package auth
 
 import (
 	"context"
+	"errors"
 	"testing"
 	"time"
 
@@ -38,6 +39,39 @@ func TestSessionStore_RegisterSameTokenTwiceIsRejected(t *testing.T) {
 	assert.ErrorIs(t, err, ErrTokenAlreadyUsed)
 }
 
+func TestSessionStore_ConcurrentRegistrationAllowsOneCaller(t *testing.T) {
+	s := newTestSessionStore(t)
+	ctx := context.Background()
+	const attempts = 100
+
+	start := make(chan struct{})
+	results := make(chan error, attempts)
+	for range attempts {
+		go func() {
+			<-start
+			results <- s.RegisterToken(ctx, "token", time.Now().Add(time.Hour))
+		}()
+	}
+	close(start)
+
+	succeeded := 0
+	alreadyUsed := 0
+	for range attempts {
+		err := <-results
+		switch {
+		case err == nil:
+			succeeded++
+		case errors.Is(err, ErrTokenAlreadyUsed):
+			alreadyUsed++
+		default:
+			require.NoError(t, err)
+		}
+	}
+
+	assert.Equal(t, 1, succeeded)
+	assert.Equal(t, attempts-1, alreadyUsed)
+}
+
 func TestSessionStore_RegisterDifferentTokensAreIndependent(t *testing.T) {
 	s := newTestSessionStore(t)
 	ctx := context.Background()
@@ -72,6 +106,23 @@ func TestSessionStore_EntryEvictsAtTTLAndAllowsReRegistration(t *testing.T) {
 	require.NoError(t, s.RegisterToken(ctx, token, time.Now().Add(time.Hour)))
 }
 
+type failingTokenCache struct {
+	err error
+}
+
+func (f failingTokenCache) SetNX(context.Context, string, string, time.Duration) (bool, error) {
+	return false, f.err
+}
+
+func TestSessionStore_CacheErrorIsReturned(t *testing.T) {
+	cacheErr := errors.New("cache unavailable")
+	s := NewSessionStore(failingTokenCache{err: cacheErr})
+
+	err := s.RegisterToken(context.Background(), "token", time.Now().Add(time.Hour))
+	require.Error(t, err)
+	assert.ErrorIs(t, err, cacheErr)
+}
+
 func TestHashToken_StableAndDoesNotLeak(t *testing.T) {
 	a := hashToken("tokenA")
 	b := hashToken("tokenB")

management/server/cache/memory.go

@@ -0,0 +1,31 @@
+package cache
+
+import (
+	"context"
+	"time"
+
+	"github.com/eko/gocache/lib/v4/store"
+	gocachestore "github.com/eko/gocache/store/go_cache/v4"
+	gocache "github.com/patrickmn/go-cache"
+)
+
+type goCacheStore struct {
+	store.StoreInterface
+	client *gocache.Cache
+}
+
+func newMemoryStore(maxTimeout, cleanupInterval time.Duration) Store {
+	client := gocache.New(maxTimeout, cleanupInterval)
+	return &goCacheStore{
+		StoreInterface: gocachestore.NewGoCache(client),
+		client:         client,
+	}
+}
+
+func (s *goCacheStore) SetNX(_ context.Context, key, value string, ttl time.Duration) (bool, error) {
+	// Add only returns an error when a non-expired entry already exists.
+	if err := s.client.Add(key, value, ttl); err != nil {
+		return false, nil
+	}
+	return true, nil
+}

management/server/cache/memory_test.go

@@ -0,0 +1,49 @@
+package cache_test
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/netbirdio/netbird/management/server/cache"
+)
+
+func TestMemoryStore(t *testing.T) {
+	memStore, err := cache.NewStore(context.Background(), 100*time.Millisecond, 300*time.Millisecond, 100)
+	if err != nil {
+		t.Fatalf("couldn't create memory store: %s", err)
+	}
+	ctx := context.Background()
+	key, value := "testing", "tested"
+	err = memStore.Set(ctx, key, value)
+	if err != nil {
+		t.Errorf("couldn't set testing data: %s", err)
+	}
+	result, err := memStore.Get(ctx, key)
+	if err != nil {
+		t.Errorf("couldn't get testing data: %s", err)
+	}
+	if value != result.(string) {
+		t.Errorf("value returned doesn't match testing data, got %s, expected %s", result, value)
+	}
+	created, err := memStore.SetNX(ctx, "atomic", value, 100*time.Millisecond)
+	if err != nil {
+		t.Fatalf("couldn't atomically set testing data: %s", err)
+	}
+	if !created {
+		t.Fatal("first atomic set should create the entry")
+	}
+	created, err = memStore.SetNX(ctx, "atomic", value, 100*time.Millisecond)
+	if err != nil {
+		t.Fatalf("couldn't atomically check testing data: %s", err)
+	}
+	if created {
+		t.Fatal("second atomic set should not replace the entry")
+	}
+	// test expiration
+	time.Sleep(300 * time.Millisecond)
+	_, err = memStore.Get(ctx, key)
+	if err == nil {
+		t.Error("value should not be found")
+	}
+}

management/server/cache/redis.go

@@ -0,0 +1,51 @@
+package cache
+
+import (
+	"context"
+	"fmt"
+	"math"
+	"time"
+
+	"github.com/eko/gocache/lib/v4/store"
+	redisstore "github.com/eko/gocache/store/redis/v4"
+	"github.com/redis/go-redis/v9"
+	log "github.com/sirupsen/logrus"
+)
+
+type redisStore struct {
+	store.StoreInterface
+	client *redis.Client
+}
+
+func getRedisStore(ctx context.Context, redisEnvAddr string, maxConn int) (Store, error) {
+	options, err := redis.ParseURL(redisEnvAddr)
+	if err != nil {
+		return nil, fmt.Errorf("parsing redis cache url: %s", err)
+	}
+
+	options.MaxIdleConns = int(math.Ceil(float64(maxConn) * 0.5)) // 50% of max conns
+	options.MinIdleConns = int(math.Ceil(float64(maxConn) * 0.1)) // 10% of max conns
+	options.MaxActiveConns = maxConn
+	options.ConnMaxIdleTime = 30 * time.Minute
+	options.ConnMaxLifetime = 0
+	options.PoolTimeout = 10 * time.Second
+	redisClient := redis.NewClient(options)
+	subCtx, cancel := context.WithTimeout(ctx, 2*time.Second)
+	defer cancel()
+
+	_, err = redisClient.Ping(subCtx).Result()
+	if err != nil {
+		return nil, err
+	}
+
+	log.WithContext(subCtx).Infof("using redis cache at %s", redisEnvAddr)
+
+	return &redisStore{
+		StoreInterface: redisstore.NewRedis(redisClient),
+		client:         redisClient,
+	}, nil
+}
+
+func (s *redisStore) SetNX(ctx context.Context, key, value string, ttl time.Duration) (bool, error) {
+	return s.client.SetNX(ctx, key, value, ttl).Result()
+}

management/server/cache/redis_test.go

@@ -12,32 +12,6 @@ import (
 	"github.com/netbirdio/netbird/management/server/cache"
 )
 
-func TestMemoryStore(t *testing.T) {
-	memStore, err := cache.NewStore(context.Background(), 100*time.Millisecond, 300*time.Millisecond, 100)
-	if err != nil {
-		t.Fatalf("couldn't create memory store: %s", err)
-	}
-	ctx := context.Background()
-	key, value := "testing", "tested"
-	err = memStore.Set(ctx, key, value)
-	if err != nil {
-		t.Errorf("couldn't set testing data: %s", err)
-	}
-	result, err := memStore.Get(ctx, key)
-	if err != nil {
-		t.Errorf("couldn't get testing data: %s", err)
-	}
-	if value != result.(string) {
-		t.Errorf("value returned doesn't match testing data, got %s, expected %s", result, value)
-	}
-	// test expiration
-	time.Sleep(300 * time.Millisecond)
-	_, err = memStore.Get(ctx, key)
-	if err == nil {
-		t.Error("value should not be found")
-	}
-}
-
 func TestRedisStoreConnectionFailure(t *testing.T) {
 	t.Setenv(cache.RedisStoreEnvVar, "redis://127.0.0.1:6379")
 	_, err := cache.NewStore(context.Background(), 10*time.Millisecond, 30*time.Millisecond, 100)
@@ -94,6 +68,47 @@ func TestRedisStoreConnectionSuccess(t *testing.T) {
 	if value != r {
 		t.Errorf("value returned from redis doesn't match testing data, got %s, expected %s", r, value)
 	}
+
+	secondRedisStore, err := cache.NewStore(context.Background(), 100*time.Millisecond, 300*time.Millisecond, 100)
+	if err != nil {
+		t.Fatalf("couldn't create second redis store: %s", err)
+	}
+	start := make(chan struct{})
+	type setResult struct {
+		created bool
+		err     error
+	}
+	results := make(chan setResult, 2)
+	for _, cacheStore := range []cache.Store{redisStore, secondRedisStore} {
+		go func() {
+			<-start
+			created, err := cacheStore.SetNX(ctx, "atomic", value, time.Second)
+			results <- setResult{created: created, err: err}
+		}()
+	}
+	close(start)
+
+	created := 0
+	for range 2 {
+		result := <-results
+		if result.err != nil {
+			t.Fatalf("atomic redis set failed: %s", result.err)
+		}
+		if result.created {
+			created++
+		}
+	}
+	if created != 1 {
+		t.Fatalf("expected exactly one redis client to create the entry, got %d", created)
+	}
+	ttl, err := redisClient.PTTL(ctx, "atomic").Result()
+	if err != nil {
+		t.Fatalf("couldn't read atomic entry TTL: %s", err)
+	}
+	if ttl <= 0 {
+		t.Fatalf("atomic entry should have a positive TTL, got %s", ttl)
+	}
+
 	// test expiration
 	time.Sleep(300 * time.Millisecond)
 	_, err = redisStore.Get(ctx, key)

management/server/cache/store.go

@@ -2,17 +2,10 @@ package cache
 
 import (
 	"context"
-	"fmt"
-	"math"
 	"os"
 	"time"
 
 	"github.com/eko/gocache/lib/v4/store"
-	gocache_store "github.com/eko/gocache/store/go_cache/v4"
-	redis_store "github.com/eko/gocache/store/redis/v4"
-	gocache "github.com/patrickmn/go-cache"
-	"github.com/redis/go-redis/v9"
-	log "github.com/sirupsen/logrus"
 )
 
 // RedisStoreEnvVar is the environment variable that determines if a redis store should be used.
@@ -31,15 +24,21 @@ const (
 	DefaultStoreMaxConn = 1000
 )
 
+// Store extends the shared cache interface with atomic insertion support.
+type Store interface {
+	store.StoreInterface
+	// SetNX atomically stores a value with a TTL only when the key does not exist.
+	SetNX(ctx context.Context, key, value string, ttl time.Duration) (bool, error)
+}
+
 // NewStore creates a new cache store with the given max timeout and cleanup interval. It checks for the environment Variable RedisStoreEnvVar
 // to determine if a redis store should be used. If the environment variable is set, it will attempt to connect to the redis store.
-func NewStore(ctx context.Context, maxTimeout, cleanupInterval time.Duration, maxConn int) (store.StoreInterface, error) {
+func NewStore(ctx context.Context, maxTimeout, cleanupInterval time.Duration, maxConn int) (Store, error) {
 	redisAddr := GetAddrFromEnv()
 	if redisAddr != "" {
 		return getRedisStore(ctx, redisAddr, maxConn)
 	}
-	goc := gocache.New(maxTimeout, cleanupInterval)
-	return gocache_store.NewGoCache(goc), nil
+	return newMemoryStore(maxTimeout, cleanupInterval), nil
 }
 
 // GetAddrFromEnv returns the redis address from the environment variable RedisStoreEnvVar or its legacy counterpart.
@@ -50,29 +49,3 @@ func GetAddrFromEnv() string {
 	}
 	return addr
 }
-
-func getRedisStore(ctx context.Context, redisEnvAddr string, maxConn int) (store.StoreInterface, error) {
-	options, err := redis.ParseURL(redisEnvAddr)
-	if err != nil {
-		return nil, fmt.Errorf("parsing redis cache url: %s", err)
-	}
-
-	options.MaxIdleConns = int(math.Ceil(float64(maxConn) * 0.5)) // 50% of max conns
-	options.MinIdleConns = int(math.Ceil(float64(maxConn) * 0.1)) // 10% of max conns
-	options.MaxActiveConns = maxConn
-	options.ConnMaxIdleTime = 30 * time.Minute
-	options.ConnMaxLifetime = 0
-	options.PoolTimeout = 10 * time.Second
-	redisClient := redis.NewClient(options)
-	subCtx, cancel := context.WithTimeout(ctx, 2*time.Second)
-	defer cancel()
-
-	_, err = redisClient.Ping(subCtx).Result()
-	if err != nil {
-		return nil, err
-	}
-
-	log.WithContext(subCtx).Infof("using redis cache at %s", redisEnvAddr)
-
-	return redis_store.NewRedis(redisClient), nil
-}