Merge pull request #1539 from hashicorp/f-reap

Adds child process reaping when Consul is running as PID 1.

command/agent/command.go

@@ -18,6 +18,7 @@ import (
 	"github.com/armon/go-metrics/datadog"
 	"github.com/hashicorp/consul/watch"
 	"github.com/hashicorp/go-checkpoint"
+	"github.com/hashicorp/go-reap"
 	"github.com/hashicorp/go-syslog"
 	"github.com/hashicorp/logutils"
 	scada "github.com/hashicorp/scada-client"
@@ -641,6 +642,33 @@ func (c *Command) Run(args []string) int {
 		defer server.Shutdown()
 	}
 
+	// Enable child process reaping
+	if (config.Reap != nil && *config.Reap) || (config.Reap == nil && os.Getpid() == 1) {
+		if !reap.IsSupported() {
+			c.Ui.Error("Child process reaping is not supported on this platform (set reap=false)")
+			return 1
+		} else {
+			logger := c.agent.logger
+			logger.Printf("[DEBUG] Automatically reaping child processes")
+
+			pids := make(reap.PidCh, 1)
+			errors := make(reap.ErrorCh, 1)
+			go func() {
+				for {
+					select {
+					case pid := <-pids:
+						logger.Printf("[DEBUG] Reaped child process %d", pid)
+					case err := <-errors:
+						logger.Printf("[ERR] Error reaping child process: %v", err)
+					case <-c.agent.shutdownCh:
+						return
+					}
+				}
+			}()
+			go reap.ReapChildren(pids, errors, c.agent.shutdownCh)
+		}
+	}
+
 	// Check and shut down the SCADA listeners at the end
 	defer func() {
 		if c.scadaHttp != nil {

command/agent/config.go

@@ -422,6 +422,17 @@ type Config struct {
 	// Minimum Session TTL
 	SessionTTLMin    time.Duration `mapstructure:"-"`
 	SessionTTLMinRaw string        `mapstructure:"session_ttl_min"`
+
+	// Reap controls automatic reaping of child processes, useful if running
+	// as PID 1 in a Docker container. This defaults to nil which will make
+	// Consul reap only if it detects it's running as PID 1. If non-nil,
+	// then this will be used to decide if reaping is enabled.
+	Reap *bool `mapstructure:"reap"`
+}
+
+// Bool is used to initialize bool pointers in struct literals.
+func Bool(b bool) *bool {
+	return &b
 }
 
 // UnixSocketPermissions contains information about a unix socket, and
@@ -1140,6 +1151,10 @@ func MergeConfig(a, b *Config) *Config {
 	result.RetryJoinWan = append(result.RetryJoinWan, a.RetryJoinWan...)
 	result.RetryJoinWan = append(result.RetryJoinWan, b.RetryJoinWan...)
 
+	if b.Reap != nil {
+		result.Reap = b.Reap
+	}
+
 	return &result
 }
 

command/agent/config_test.go

@@ -777,6 +777,27 @@ func TestDecodeConfig(t *testing.T) {
 	if config.SessionTTLMin != 5*time.Second {
 		t.Fatalf("bad: %s %#v", config.SessionTTLMin.String(), config)
 	}
+
+	// Reap
+	input = `{"reap": true}`
+	config, err = DecodeConfig(bytes.NewReader([]byte(input)))
+	if err != nil {
+		t.Fatalf("err: %s", err)
+	}
+
+	if config.Reap == nil || *config.Reap != true {
+		t.Fatalf("bad: reap not enabled: %#v", config)
+	}
+
+	input = `{}`
+	config, err = DecodeConfig(bytes.NewReader([]byte(input)))
+	if err != nil {
+		t.Fatalf("err: %s", err)
+	}
+
+	if config.Reap != nil {
+		t.Fatalf("bad: reap not tri-stated: %#v", config)
+	}
 }
 
 func TestDecodeConfig_invalidKeys(t *testing.T) {
@@ -1266,6 +1287,7 @@ func TestMergeConfig(t *testing.T) {
 			RPC:        &net.TCPAddr{},
 			RPCRaw:     "127.0.0.5:1233",
 		},
+		Reap: Bool(true),
 	}
 
 	c := MergeConfig(a, b)

website/source/docs/agent/options.html.markdown

@@ -504,6 +504,11 @@ definitions support being updated during a reload.
 * <a name="protocol"></a><a href="#protocol">`protocol`</a> Equivalent to the
   [`-protocol` command-line flag](#_protocol).
 
+* <a name="reap"></a><a href="#reap">`reap`</a> controls Consul's automatic reaping of child processes, which
+  is useful if Consul is running as PID 1 in a Docker container. If this isn't specified, then Consul will
+  automatically reap child processes if it detects it is running as PID 1. If this is specified, then it
+  controls reaping regardless of Consul's PID.
+
 * <a name="recursor"></a><a href="#recursor">`recursor`</a> Provides a single recursor address.
   This has been deprecated, and the value is appended to the [`recursors`](#recursors) list for
   backwards compatibility.