feat(shutdown): passive drain and rename DisableSoftCrash

Shutdown now proceeds in three phases: close listeners immediately on signal, broadcast the in-game countdown, wait up to ShutdownDrainSeconds (default 30) for sessions to disconnect naturally via DrainPassive, then force-close any stragglers. This prevents players from entering new quests after the countdown starts, and lets mid-quest sessions finish saving without being killed mid-write. A second SIGINT during passive drain cancels it so the force-close phase runs immediately. Renamed DisableSoftCrash -> DisableShutdownCountdown since the flag controls the countdown, not crash behaviour. Existing config.json files keep working via a Viper alias on the legacy key. Closes #179.
2026-05-06 06:13:50 +02:00 · 2026-04-06 19:32:35 +02:00
parent 9b0f735335
commit e48d33ca76
7 changed files with 134 additions and 64 deletions
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

 ## [Unreleased]

+### Changed
+
+- Shutdown now proceeds in three phases: listeners close immediately on signal, then a passive drain waits up to `ShutdownDrainSeconds` (default 30) for sessions to disconnect naturally, then remaining connections are force-closed. This prevents players from starting new quests after the countdown begins ([#179](https://github.com/Mezeporta/Erupe/issues/179)).
+- Renamed config field `DisableSoftCrash` → `DisableShutdownCountdown` for clarity. The old key is still accepted via a Viper alias so existing `config.json` files keep working without modification ([#179](https://github.com/Mezeporta/Erupe/issues/179)).
+
 ### Added

 - Reverse-engineered user binary data types from `mhfo-hd.dll` via Ghidra: type 1 = character name (max 17B SJIS), type 2 = player profile with self-introduction (208B), type 3 = equipment/appearance snapshot (384B). Added structured parsing with size validation warnings to `handleMsgSysSetUserBinary`.
--- a/config.reference.json
+++ b/config.reference.json
@@ -2,7 +2,9 @@
  "Host": "127.0.0.1",
  "BinPath": "bin",
  "Language": "en",
-  "DisableSoftCrash": false,
+  "DisableShutdownCountdown": false,
+  "ShutdownCountdownSeconds": 10,
+  "ShutdownDrainSeconds": 30,
  "HideLoginNotice": true,
  "LoginNotices": [
    "<BODY><CENTER><SIZE_3><C_4>Welcome to Erupe SU9.3!<BR><BODY><LEFT><SIZE_2><C_5>Erupe is experimental software<C_7>, we are not liable for any<BR><BODY>issues caused by installing the software!<BR><BODY><BR><BODY><C_4>■Report bugs on Discord!<C_7><BR><BODY><BR><BODY><C_4>■Test everything!<C_7><BR><BODY><BR><BODY><C_4>■Don't talk to softlocking NPCs!<C_7><BR><BODY><BR><BODY><C_4>■Fork the code on GitHub!<C_7><BR><BODY><BR><BODY>Thank you to all of the contributors,<BR><BODY><BR><BODY>this wouldn't exist without you."
--- a/config/config.go
+++ b/config/config.go
@@ -64,30 +64,31 @@ func (m Mode) String() string {

 // Config holds the global server-wide config.
 type Config struct {
-	Host                   string `mapstructure:"Host"`
-	BinPath                string `mapstructure:"BinPath"`
-	Language               string
-	DisableSoftCrash            bool // Disables the 'Press Return to exit' dialog allowing scripts to reboot the server automatically
-	ShutdownCountdownSeconds    int  // Seconds to count down before shutting down (default 10; ignored when DisableSoftCrash is true)
-	HideLoginNotice        bool     // Hide the Erupe notice on login
-	LoginNotices           []string // MHFML string of the login notices displayed
-	PatchServerManifest    string   // Manifest patch server override
-	PatchServerFile        string   // File patch server override
-	DeleteOnSaveCorruption    bool // Attempts to save corrupted data will flag the save for deletion
-	DisableSaveIntegrityCheck bool // Skip SHA-256 hash verification on load (needed for cross-server save transfers)
-	ClientMode             string
-	RealClientMode         Mode
-	QuestCacheExpiry       int    // Number of seconds to keep quest data cached
-	CommandPrefix          string // The prefix for commands
-	AutoCreateAccount      bool   // Automatically create accounts if they don't exist
-	LoopDelay              int    // Delay in milliseconds between each loop iteration
-	DefaultCourses         []uint16
-	EarthStatus            int32
-	EarthID                int32
-	EarthMonsters          []int32
-	SaveDumps              SaveDumpOptions
-	Screenshots            ScreenshotsOptions
-	Capture                CaptureOptions
+	Host                      string `mapstructure:"Host"`
+	BinPath                   string `mapstructure:"BinPath"`
+	Language                  string
+	DisableShutdownCountdown  bool     `mapstructure:"DisableShutdownCountdown"` // Skip the in-game shutdown countdown (lets scripts restart the server unattended). Previously named DisableSoftCrash — legacy key still accepted.
+	ShutdownCountdownSeconds  int      // Seconds to count down before shutting down (default 10; ignored when DisableShutdownCountdown is true)
+	ShutdownDrainSeconds      int      // Additional seconds to wait for sessions to disconnect naturally before force-closing (default 30)
+	HideLoginNotice           bool     // Hide the Erupe notice on login
+	LoginNotices              []string // MHFML string of the login notices displayed
+	PatchServerManifest       string   // Manifest patch server override
+	PatchServerFile           string   // File patch server override
+	DeleteOnSaveCorruption    bool     // Attempts to save corrupted data will flag the save for deletion
+	DisableSaveIntegrityCheck bool     // Skip SHA-256 hash verification on load (needed for cross-server save transfers)
+	ClientMode                string
+	RealClientMode            Mode
+	QuestCacheExpiry          int    // Number of seconds to keep quest data cached
+	CommandPrefix             string // The prefix for commands
+	AutoCreateAccount         bool   // Automatically create accounts if they don't exist
+	LoopDelay                 int    // Delay in milliseconds between each loop iteration
+	DefaultCourses            []uint16
+	EarthStatus               int32
+	EarthID                   int32
+	EarthMonsters             []int32
+	SaveDumps                 SaveDumpOptions
+	Screenshots               ScreenshotsOptions
+	Capture                   CaptureOptions

 	DebugOptions    DebugOptions
 	GameplayOptions GameplayOptions
@@ -356,6 +357,11 @@ func registerDefaults() {
 	viper.SetDefault("AutoCreateAccount", true)
 	viper.SetDefault("LoopDelay", 50)
 	viper.SetDefault("ShutdownCountdownSeconds", 10)
+	viper.SetDefault("ShutdownDrainSeconds", 30)
+	// Back-compat: old configs use DisableSoftCrash. RegisterAlias makes Viper
+	// treat reads/writes of the old key as the new key, so existing
+	// config.json files keep working without modification.
+	viper.RegisterAlias("DisableSoftCrash", "DisableShutdownCountdown")
 	viper.SetDefault("DefaultCourses", []uint16{1, 23, 24})
 	viper.SetDefault("EarthMonsters", []int32{0, 0, 0, 0})

--- a/config/config_load_test.go
+++ b/config/config_load_test.go
@@ -142,26 +142,26 @@ func TestLoadConfigDefaultModeWhenInvalid(t *testing.T) {
 // TestConfigStruct tests Config structure creation with all fields
 func TestConfigStruct(t *testing.T) {
 	cfg := &Config{
-		Host:                   "localhost",
-		BinPath:                "/opt/erupe",
-		Language:               "en",
-		DisableSoftCrash:       false,
-		HideLoginNotice:        false,
-		LoginNotices:           []string{"Welcome"},
-		PatchServerManifest:    "http://patch.example.com/manifest",
-		PatchServerFile:        "http://patch.example.com/files",
+		Host:                      "localhost",
+		BinPath:                   "/opt/erupe",
+		Language:                  "en",
+		DisableShutdownCountdown:  false,
+		HideLoginNotice:           false,
+		LoginNotices:              []string{"Welcome"},
+		PatchServerManifest:       "http://patch.example.com/manifest",
+		PatchServerFile:           "http://patch.example.com/files",
 		DeleteOnSaveCorruption:    false,
 		DisableSaveIntegrityCheck: false,
-		ClientMode:             "ZZ",
-		RealClientMode:         ZZ,
-		QuestCacheExpiry:       3600,
-		CommandPrefix:          "!",
-		AutoCreateAccount:      false,
-		LoopDelay:              100,
-		DefaultCourses:         []uint16{1, 2, 3},
-		EarthStatus:            0,
-		EarthID:                0,
-		EarthMonsters:          []int32{100, 101, 102},
+		ClientMode:                "ZZ",
+		RealClientMode:            ZZ,
+		QuestCacheExpiry:          3600,
+		CommandPrefix:             "!",
+		AutoCreateAccount:         false,
+		LoopDelay:                 100,
+		DefaultCourses:            []uint16{1, 2, 3},
+		EarthStatus:               0,
+		EarthID:                   0,
+		EarthMonsters:             []int32{100, 101, 102},
 		SaveDumps: SaveDumpOptions{
 			Enabled:    true,
 			RawEnabled: false,
--- a/config/config_test.go
+++ b/config/config_test.go
@@ -149,26 +149,26 @@ func TestGetOutboundIP4(t *testing.T) {
 // TestConfigStructTypes verifies Config struct fields have correct types
 func TestConfigStructTypes(t *testing.T) {
 	cfg := &Config{
-		Host:                   "localhost",
-		BinPath:                "/path/to/bin",
-		Language:               "en",
-		DisableSoftCrash:       false,
-		HideLoginNotice:        false,
-		LoginNotices:           []string{"Notice"},
-		PatchServerManifest:    "http://patch.example.com",
-		PatchServerFile:        "http://files.example.com",
+		Host:                      "localhost",
+		BinPath:                   "/path/to/bin",
+		Language:                  "en",
+		DisableShutdownCountdown:  false,
+		HideLoginNotice:           false,
+		LoginNotices:              []string{"Notice"},
+		PatchServerManifest:       "http://patch.example.com",
+		PatchServerFile:           "http://files.example.com",
 		DeleteOnSaveCorruption:    false,
 		DisableSaveIntegrityCheck: false,
-		ClientMode:             "ZZ",
-		RealClientMode:         ZZ,
-		QuestCacheExpiry:       3600,
-		CommandPrefix:          "!",
-		AutoCreateAccount:      false,
-		LoopDelay:              100,
-		DefaultCourses:         []uint16{1, 2, 3},
-		EarthStatus:            1,
-		EarthID:                1,
-		EarthMonsters:          []int32{1, 2, 3},
+		ClientMode:                "ZZ",
+		RealClientMode:            ZZ,
+		QuestCacheExpiry:          3600,
+		CommandPrefix:             "!",
+		AutoCreateAccount:         false,
+		LoopDelay:                 100,
+		DefaultCourses:            []uint16{1, 2, 3},
+		EarthStatus:               1,
+		EarthID:                   1,
+		EarthMonsters:             []int32{1, 2, 3},
 		SaveDumps: SaveDumpOptions{
 			Enabled:    true,
 			RawEnabled: false,
--- a/main.go
+++ b/main.go
@@ -14,8 +14,8 @@ import (
 	"syscall"
 	"time"

-	cfg "erupe-ce/config"
 	"erupe-ce/common/gametime"
+	cfg "erupe-ce/config"
 	"erupe-ce/server/api"
 	"erupe-ce/server/channelserver"
 	"erupe-ce/server/discordbot"
@@ -401,7 +401,15 @@ func main() {
 	signal.Notify(sig, os.Interrupt, syscall.SIGTERM)
 	<-sig

-	if !config.DisableSoftCrash {
+	// Phase 1: stop accepting new connections immediately so players seeing
+	// the countdown cannot start fresh quests at T-1.
+	if config.Channel.Enabled {
+		for _, c := range channels {
+			c.Shutdown()
+		}
+	}
+
+	if !config.DisableShutdownCountdown {
 		countdown := config.ShutdownCountdownSeconds
 		if countdown <= 0 {
 			countdown = 10
@@ -422,6 +430,31 @@ func main() {
 	}

 	if config.Channel.Enabled {
+		// Phase 2: passive drain — give active sessions (mid-quest players)
+		// up to ShutdownDrainSeconds to disconnect on their own.
+		drainSecs := config.ShutdownDrainSeconds
+		if drainSecs < 0 {
+			drainSecs = 0
+		}
+		if drainSecs > 0 {
+			passiveCtx, passiveCancel := context.WithTimeout(context.Background(), time.Duration(drainSecs)*time.Second)
+			// A second signal cancels the passive drain so the force-close
+			// phase runs immediately.
+			go func() {
+				select {
+				case <-sig:
+					logger.Info("Second signal received, skipping passive drain")
+					passiveCancel()
+				case <-passiveCtx.Done():
+				}
+			}()
+			for _, c := range channels {
+				c.DrainPassive(passiveCtx)
+			}
+			passiveCancel()
+		}
+
+		// Phase 3: force-close any stragglers so logoutPlayer runs and saves.
 		drainCtx, drainCancel := context.WithTimeout(context.Background(), 30*time.Second)
 		defer drainCancel()
 		for _, c := range channels {
@@ -451,7 +484,7 @@ func wait() {
 }

 func preventClose(config *cfg.Config, text string) {
-	if config != nil && config.DisableSoftCrash {
+	if config != nil && config.DisableShutdownCountdown {
 		os.Exit(0)
 	}
 	fmt.Println("\nFailed to start Erupe:\n" + text)
--- a/server/channelserver/sys_channel_server.go
+++ b/server/channelserver/sys_channel_server.go
@@ -249,6 +249,30 @@ func (s *Server) Shutdown() {

 }

+// DrainPassive waits for active sessions to disconnect naturally (e.g. players
+// finishing a quest and logging out) without force-closing anything. Returns
+// when the session count reaches zero or ctx is cancelled. Callers should have
+// already invoked Shutdown() to close the listener so no new sessions arrive.
+func (s *Server) DrainPassive(ctx context.Context) {
+	ticker := time.NewTicker(100 * time.Millisecond)
+	defer ticker.Stop()
+	for {
+		s.Lock()
+		n := len(s.sessions)
+		s.Unlock()
+		if n == 0 {
+			s.logger.Info("Passive drain complete: all sessions disconnected")
+			return
+		}
+		select {
+		case <-ctx.Done():
+			s.logger.Info("Passive drain deadline reached", zap.Int("remaining_sessions", n))
+			return
+		case <-ticker.C:
+		}
+	}
+}
+
 // ShutdownAndDrain stops accepting new connections, force-closes every active
 // session so that their logoutPlayer cleanup runs (saves character data, removes
 // from stages, etc.), then waits until all sessions have been removed from the