Add config parameter to change stop timeout during daemon shutdown

This fix tries to add a daemon config parameter `--shutdown-timeout`
that specifies the timeout value to stop containers gracefully
(before SIGKILL). The default value is 15s.

The `--shutdown-timeout` parameter is added to daemon options and
config file. It will also be updated during daemon reload.

Additional test cases have been added to cover the change.

This fix fixes #22471.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

daemon/config.go

@@ -37,6 +37,10 @@ const (
 	disableNetworkBridge = "none"
 )
 
+const (
+	defaultShutdownTimeout = 15
+)
+
 // flatOptions contains configuration keys
 // that MUST NOT be parsed as deep structures.
 // Use this to differentiate these options
@@ -123,6 +127,10 @@ type CommonConfig struct {
 	// may take place at a time for each push.
 	MaxConcurrentUploads *int `json:"max-concurrent-uploads,omitempty"`
 
+	// ShutdownTimeout is the timeout value (in seconds) the daemon will wait for the container
+	// to stop when daemon is being shutdown
+	ShutdownTimeout int `json:"shutdown-timeout,omitempty"`
+
 	Debug     bool     `json:"debug,omitempty"`
 	Hosts     []string `json:"hosts,omitempty"`
 	LogLevel  string   `json:"log-level,omitempty"`
@@ -176,6 +184,7 @@ func (config *Config) InstallCommonFlags(flags *pflag.FlagSet) {
 	flags.StringVar(&config.CorsHeaders, "api-cors-header", "", "Set CORS headers in the remote API")
 	flags.IntVar(&maxConcurrentDownloads, "max-concurrent-downloads", defaultMaxConcurrentDownloads, "Set the max concurrent downloads for each pull")
 	flags.IntVar(&maxConcurrentUploads, "max-concurrent-uploads", defaultMaxConcurrentUploads, "Set the max concurrent uploads for each push")
+	flags.IntVar(&config.ShutdownTimeout, "shutdown-timeout", defaultShutdownTimeout, "Set the default shutdown timeout")
 
 	flags.StringVar(&config.SwarmDefaultAdvertiseAddr, "swarm-default-advertise-addr", "", "Set default address or interface for swarm advertised address")
 

daemon/daemon.go

@@ -732,12 +732,13 @@ func (daemon *Daemon) shutdownContainer(c *container.Container) error {
 	return nil
 }
 
-// ShutdownTimeout returns the shutdown timeout based on the max stopTimeout of the containers
+// ShutdownTimeout returns the shutdown timeout based on the max stopTimeout of the containers,
+// and is limited by daemon's ShutdownTimeout.
 func (daemon *Daemon) ShutdownTimeout() int {
-	// By default we use container.DefaultStopTimeout + 5s, which is 15s.
-	// TODO (yongtang): Will need to allow shutdown-timeout once #23036 is in place.
+	// By default we use daemon's ShutdownTimeout.
+	shutdownTimeout := daemon.configStore.ShutdownTimeout
+
 	graceTimeout := 5
-	shutdownTimeout := container.DefaultStopTimeout + graceTimeout
 	if daemon.containers != nil {
 		for _, c := range daemon.containers.List() {
 			if shutdownTimeout >= 0 {
@@ -769,7 +770,7 @@ func (daemon *Daemon) Shutdown() error {
 	}
 
 	if daemon.containers != nil {
-		logrus.Debug("starting clean shutdown of all containers...")
+		logrus.Debugf("start clean shutdown of all containers with a %d seconds timeout...", daemon.configStore.ShutdownTimeout)
 		daemon.containers.ApplyAll(func(c *container.Container) {
 			if !c.IsRunning() {
 				return
@@ -995,6 +996,7 @@ func (daemon *Daemon) initDiscovery(config *Config) error {
 // - Daemon max concurrent uploads
 // - Cluster discovery (reconfigure and restart).
 // - Daemon live restore
+// - Daemon shutdown timeout (in seconds).
 func (daemon *Daemon) Reload(config *Config) (err error) {
 
 	daemon.configStore.reloadLock.Lock()
@@ -1055,6 +1057,11 @@ func (daemon *Daemon) Reload(config *Config) (err error) {
 		daemon.uploadManager.SetConcurrency(*daemon.configStore.MaxConcurrentUploads)
 	}
 
+	if config.IsValueSet("shutdown-timeout") {
+		daemon.configStore.ShutdownTimeout = config.ShutdownTimeout
+		logrus.Debugf("Reset Shutdown Timeout: %d", daemon.configStore.ShutdownTimeout)
+	}
+
 	// We emit daemon reload event here with updatable configurations
 	attributes["debug"] = fmt.Sprintf("%t", daemon.configStore.Debug)
 	attributes["live-restore"] = fmt.Sprintf("%t", daemon.configStore.LiveRestoreEnabled)
@@ -1074,6 +1081,7 @@ func (daemon *Daemon) Reload(config *Config) (err error) {
 	}
 	attributes["max-concurrent-downloads"] = fmt.Sprintf("%d", *daemon.configStore.MaxConcurrentDownloads)
 	attributes["max-concurrent-uploads"] = fmt.Sprintf("%d", *daemon.configStore.MaxConcurrentUploads)
+	attributes["shutdown-timeout"] = fmt.Sprintf("%d", daemon.configStore.ShutdownTimeout)
 
 	return nil
 }

docs/reference/commandline/dockerd.md

@@ -64,6 +64,7 @@ Options:
       --raw-logs                              Full timestamps without ANSI coloring
       --registry-mirror value                 Preferred Docker registry mirror (default [])
       --selinux-enabled                       Enable selinux support
+      --shutdown-timeout=15                   Set the shutdown timeout value in seconds
   -s, --storage-driver string                 Storage driver to use
       --storage-opt value                     Storage driver options (default [])
       --swarm-default-advertise-addr string   Set default address or interface for swarm advertised address
@@ -1118,6 +1119,7 @@ This is a full example of the allowed configuration options on Linux:
 	"cluster-advertise": "",
 	"max-concurrent-downloads": 3,
 	"max-concurrent-uploads": 5,
+	"shutdown-timeout": 15,
 	"debug": true,
 	"hosts": [],
 	"log-level": "",
@@ -1194,6 +1196,7 @@ This is a full example of the allowed configuration options on Windows:
     "graph": "",
     "cluster-store": "",
     "cluster-advertise": "",
+    "shutdown-timeout": 15,
     "debug": true,
     "hosts": [],
     "log-level": "",

integration-cli/docker_cli_daemon_test.go

@@ -2920,3 +2920,57 @@ func (s *DockerDaemonSuite) TestDaemonWithUserlandProxyPath(c *check.C) {
 	c.Assert(out, checker.Contains, "driver failed programming external connectivity on endpoint")
 	c.Assert(out, checker.Contains, "/does/not/exist: no such file or directory")
 }
+
+// Test case for #22471
+func (s *DockerDaemonSuite) TestDaemonShutdownTimeout(c *check.C) {
+	testRequires(c, SameHostDaemon)
+
+	c.Assert(s.d.StartWithBusybox("--shutdown-timeout=3"), check.IsNil)
+
+	_, err := s.d.Cmd("run", "-d", "busybox", "top")
+	c.Assert(err, check.IsNil)
+
+	syscall.Kill(s.d.cmd.Process.Pid, syscall.SIGINT)
+
+	select {
+	case <-s.d.wait:
+	case <-time.After(5 * time.Second):
+	}
+
+	expectedMessage := `level=debug msg="start clean shutdown of all containers with a 3 seconds timeout..."`
+	content, _ := ioutil.ReadFile(s.d.logFile.Name())
+	c.Assert(string(content), checker.Contains, expectedMessage)
+}
+
+// Test case for #22471
+func (s *DockerDaemonSuite) TestDaemonShutdownTimeoutWithConfigFile(c *check.C) {
+	testRequires(c, SameHostDaemon)
+
+	// daemon config file
+	configFilePath := "test.json"
+	configFile, err := os.Create(configFilePath)
+	c.Assert(err, checker.IsNil)
+	defer os.Remove(configFilePath)
+
+	daemonConfig := `{ "shutdown-timeout" : 8 }`
+	fmt.Fprintf(configFile, "%s", daemonConfig)
+	configFile.Close()
+	c.Assert(s.d.Start(fmt.Sprintf("--config-file=%s", configFilePath)), check.IsNil)
+
+	configFile, err = os.Create(configFilePath)
+	c.Assert(err, checker.IsNil)
+	daemonConfig = `{ "shutdown-timeout" : 5 }`
+	fmt.Fprintf(configFile, "%s", daemonConfig)
+	configFile.Close()
+
+	syscall.Kill(s.d.cmd.Process.Pid, syscall.SIGHUP)
+
+	select {
+	case <-s.d.wait:
+	case <-time.After(3 * time.Second):
+	}
+
+	expectedMessage := `level=debug msg="Reset Shutdown Timeout: 5"`
+	content, _ := ioutil.ReadFile(s.d.logFile.Name())
+	c.Assert(string(content), checker.Contains, expectedMessage)
+}

integration-cli/docker_cli_events_unix_test.go

@@ -418,7 +418,7 @@ func (s *DockerDaemonSuite) TestDaemonEvents(c *check.C) {
 
 	configFile, err = os.Create(configFilePath)
 	c.Assert(err, checker.IsNil)
-	daemonConfig = `{"max-concurrent-downloads":1,"labels":["bar=foo"]}`
+	daemonConfig = `{"max-concurrent-downloads":1,"labels":["bar=foo"], "shutdown-timeout": 10}`
 	fmt.Fprintf(configFile, "%s", daemonConfig)
 	configFile.Close()
 
@@ -429,7 +429,7 @@ func (s *DockerDaemonSuite) TestDaemonEvents(c *check.C) {
 	out, err = s.d.Cmd("events", "--since=0", "--until", daemonUnixTime(c))
 	c.Assert(err, checker.IsNil)
 
-	c.Assert(out, checker.Contains, fmt.Sprintf("daemon reload %s (cluster-advertise=, cluster-store=, cluster-store-opts={}, debug=true, default-runtime=runc, labels=[\"bar=foo\"], live-restore=false, max-concurrent-downloads=1, max-concurrent-uploads=5, name=%s, runtimes=runc:{docker-runc []})", daemonID, daemonName))
+	c.Assert(out, checker.Contains, fmt.Sprintf("daemon reload %s (cluster-advertise=, cluster-store=, cluster-store-opts={}, debug=true, default-runtime=runc, labels=[\"bar=foo\"], live-restore=false, max-concurrent-downloads=1, max-concurrent-uploads=5, name=%s, runtimes=runc:{docker-runc []}, shutdown-timeout=10)", daemonID, daemonName))
 }
 
 func (s *DockerDaemonSuite) TestDaemonEventsWithFilters(c *check.C) {

man/dockerd.8.md

@@ -56,6 +56,7 @@ dockerd - Enable daemon mode
 [**--registry-mirror**[=*[]*]]
 [**-s**|**--storage-driver**[=*STORAGE-DRIVER*]]
 [**--selinux-enabled**]
+[**--shutdown-timeout**[=*15*]]
 [**--storage-opt**[=*[]*]]
 [**--swarm-default-advertise-addr**[=*IP|INTERFACE*]]
 [**--tls**]
@@ -246,6 +247,9 @@ output otherwise.
 **--selinux-enabled**=*true*|*false*
   Enable selinux support. Default is false.
 
+**--shutdown-timeout**=*15*
+  Set the shutdown timeout value in seconds. Default is `15`.
+
 **--storage-opt**=[]
   Set storage driver options. See STORAGE DRIVER OPTIONS.