package app import ( "context" "database/sql" "errors" "fmt" "log/slog" "time" "galaxy/postgres" "galaxy/redisconn" "galaxy/gamemaster/internal/adapters/postgres/migrations" "galaxy/gamemaster/internal/api/internalhttp" "galaxy/gamemaster/internal/config" "galaxy/gamemaster/internal/telemetry" "github.com/redis/go-redis/v9" ) // Runtime owns the runnable Game Master process plus the cleanup // functions that release runtime resources after shutdown. type Runtime struct { cfg config.Config app *App wiring *wiring internalServer *internalhttp.Server cleanupFns []func() error } // NewRuntime constructs the runnable Game Master process from cfg. // // The runtime opens one shared `*redis.Client`, one `*sql.DB`, and one // OpenTelemetry runtime; all are released in reverse construction order // on shutdown. Embedded goose migrations apply synchronously after the // PostgreSQL pool is opened and pinged, before any listener is constructed. func NewRuntime(ctx context.Context, cfg config.Config, logger *slog.Logger) (*Runtime, error) { if ctx == nil { return nil, errors.New("new gamemaster runtime: nil context") } if err := cfg.Validate(); err != nil { return nil, fmt.Errorf("new gamemaster runtime: %w", err) } if logger == nil { logger = slog.Default() } runtime := &Runtime{ cfg: cfg, } cleanupOnError := func(err error) (*Runtime, error) { if cleanupErr := runtime.Close(); cleanupErr != nil { return nil, fmt.Errorf("%w; cleanup: %w", err, cleanupErr) } return nil, err } telemetryRuntime, err := telemetry.NewProcess(ctx, telemetry.ProcessConfig{ ServiceName: cfg.Telemetry.ServiceName, TracesExporter: cfg.Telemetry.TracesExporter, MetricsExporter: cfg.Telemetry.MetricsExporter, TracesProtocol: cfg.Telemetry.TracesProtocol, MetricsProtocol: cfg.Telemetry.MetricsProtocol, StdoutTracesEnabled: cfg.Telemetry.StdoutTracesEnabled, StdoutMetricsEnabled: cfg.Telemetry.StdoutMetricsEnabled, }, logger) if err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: telemetry: %w", err)) } runtime.cleanupFns = append(runtime.cleanupFns, func() error { shutdownCtx, cancel := context.WithTimeout(context.Background(), cfg.ShutdownTimeout) defer cancel() return telemetryRuntime.Shutdown(shutdownCtx) }) redisClient := newRedisClient(cfg.Redis) if err := instrumentRedisClient(redisClient, telemetryRuntime); err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: %w", err)) } runtime.cleanupFns = append(runtime.cleanupFns, func() error { err := redisClient.Close() if errors.Is(err, redis.ErrClosed) { return nil } return err }) if err := pingRedis(ctx, cfg.Redis, redisClient); err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: %w", err)) } pgPool, err := postgres.OpenPrimary(ctx, cfg.Postgres.Conn, postgres.WithTracerProvider(telemetryRuntime.TracerProvider()), postgres.WithMeterProvider(telemetryRuntime.MeterProvider()), ) if err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: open postgres: %w", err)) } runtime.cleanupFns = append(runtime.cleanupFns, pgPool.Close) unregisterPGStats, err := postgres.InstrumentDBStats(pgPool, postgres.WithMeterProvider(telemetryRuntime.MeterProvider()), ) if err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: instrument postgres: %w", err)) } runtime.cleanupFns = append(runtime.cleanupFns, func() error { return unregisterPGStats() }) if err := postgres.Ping(ctx, pgPool, cfg.Postgres.Conn.OperationTimeout); err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: ping postgres: %w", err)) } if err := postgres.RunMigrations(ctx, pgPool, migrations.FS(), "."); err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: run postgres migrations: %w", err)) } wiring, err := newWiring(cfg, redisClient, pgPool, time.Now, logger, telemetryRuntime) if err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: wiring: %w", err)) } runtime.wiring = wiring runtime.cleanupFns = append(runtime.cleanupFns, wiring.close) probe := newReadinessProbe(pgPool, redisClient, cfg) internalServer, err := internalhttp.NewServer(internalhttp.Config{ Addr: cfg.InternalHTTP.Addr, ReadHeaderTimeout: cfg.InternalHTTP.ReadHeaderTimeout, ReadTimeout: cfg.InternalHTTP.ReadTimeout, WriteTimeout: cfg.InternalHTTP.WriteTimeout, IdleTimeout: cfg.InternalHTTP.IdleTimeout, }, internalhttp.Dependencies{ Logger: logger, Telemetry: telemetryRuntime, Readiness: probe, RuntimeRecords: wiring.runtimeRecords, RegisterRuntime: wiring.registerRuntimeSvc, ForceNextTurn: wiring.forceNextTurnSvc, StopRuntime: wiring.stopRuntimeSvc, PatchRuntime: wiring.patchRuntimeSvc, BanishRace: wiring.banishRaceSvc, InvalidateMemberships: wiring.membershipCache, GameLiveness: wiring.livenessSvc, EngineVersions: wiring.engineVersionSvc, CommandExecute: wiring.commandExecuteSvc, PutOrders: wiring.orderPutSvc, GetReport: wiring.reportGetSvc, }) if err != nil { return cleanupOnError(fmt.Errorf("new gamemaster runtime: internal HTTP server: %w", err)) } runtime.internalServer = internalServer runtime.app = New(cfg, internalServer, wiring.schedulerTicker, wiring.healthEventsConsumer, ) return runtime, nil } // InternalServer returns the internal HTTP server owned by runtime. It is // primarily exposed for tests; production code should not depend on it. func (runtime *Runtime) InternalServer() *internalhttp.Server { if runtime == nil { return nil } return runtime.internalServer } // Run serves the internal HTTP listener until ctx is canceled or one // component fails. func (runtime *Runtime) Run(ctx context.Context) error { if ctx == nil { return errors.New("run gamemaster runtime: nil context") } if runtime == nil { return errors.New("run gamemaster runtime: nil runtime") } if runtime.app == nil { return errors.New("run gamemaster runtime: nil app") } return runtime.app.Run(ctx) } // Close releases every runtime dependency in reverse construction order. // Close is safe to call multiple times. func (runtime *Runtime) Close() error { if runtime == nil { return nil } var joined error for index := len(runtime.cleanupFns) - 1; index >= 0; index-- { if err := runtime.cleanupFns[index](); err != nil { joined = errors.Join(joined, err) } } runtime.cleanupFns = nil return joined } // readinessProbe pings every steady-state dependency the listener // guards: PostgreSQL primary and Redis master. type readinessProbe struct { pgPool *sql.DB redisClient *redis.Client postgresTimeout time.Duration redisTimeout time.Duration } func newReadinessProbe(pgPool *sql.DB, redisClient *redis.Client, cfg config.Config) *readinessProbe { return &readinessProbe{ pgPool: pgPool, redisClient: redisClient, postgresTimeout: cfg.Postgres.Conn.OperationTimeout, redisTimeout: cfg.Redis.Conn.OperationTimeout, } } // Check pings PostgreSQL and Redis. The first failing dependency aborts // the check so callers see a single, actionable error. func (probe *readinessProbe) Check(ctx context.Context) error { if err := postgres.Ping(ctx, probe.pgPool, probe.postgresTimeout); err != nil { return err } return redisconn.Ping(ctx, probe.redisClient, probe.redisTimeout) }