From 1e4aa28c9daf1ab221b67bea158a0c31d82cec67 Mon Sep 17 00:00:00 2001
From: James Phillips <james@hashicorp.com>
Date: Mon, 21 Dec 2015 21:47:35 -0800
Subject: [PATCH] Adds child process reaping when Consul is running as PID 1.

---
 command/agent/command.go                      | 27 +++++++++++++++++++
 command/agent/config.go                       | 10 +++++++
 command/agent/config_test.go                  | 13 +++++++++
 .../source/docs/agent/options.html.markdown   |  4 +++
 4 files changed, 54 insertions(+)

diff --git a/command/agent/command.go b/command/agent/command.go
index 0993ccfa1..a5f59bee9 100644
--- a/command/agent/command.go
+++ b/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 +642,32 @@ func (c *Command) Run(args []string) int {
 		defer server.Shutdown()
 	}
 
+	// Enable child process reaping
+	if !config.DisableReap && (os.Getpid() == 1) {
+		logger := c.agent.logger
+		if !reap.IsSupported() {
+			logger.Printf("[WARN] Running as PID 1 but child process reaping is not supported on this platform, disabling")
+		} else {
+			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 {
diff --git a/command/agent/config.go b/command/agent/config.go
index c03659116..50db3244f 100644
--- a/command/agent/config.go
+++ b/command/agent/config.go
@@ -422,6 +422,12 @@ type Config struct {
 	// Minimum Session TTL
 	SessionTTLMin    time.Duration `mapstructure:"-"`
 	SessionTTLMinRaw string        `mapstructure:"session_ttl_min"`
+
+	// DisableReap controls automatic reaping of child processes, useful if
+	// running as PID 1 in a Docker container. This defaults to false, and
+	// reaping will be automatically enabled if this is false and Consul's
+	// PID is 1.
+	DisableReap bool `mapstructure:"disable_reap"`
 }
 
 // UnixSocketPermissions contains information about a unix socket, and
@@ -1140,6 +1146,10 @@ func MergeConfig(a, b *Config) *Config {
 	result.RetryJoinWan = append(result.RetryJoinWan, a.RetryJoinWan...)
 	result.RetryJoinWan = append(result.RetryJoinWan, b.RetryJoinWan...)
 
+	if b.DisableReap {
+		result.DisableReap = true
+	}
+
 	return &result
 }
 
diff --git a/command/agent/config_test.go b/command/agent/config_test.go
index 000567b26..8ce8236ad 100644
--- a/command/agent/config_test.go
+++ b/command/agent/config_test.go
@@ -777,6 +777,17 @@ func TestDecodeConfig(t *testing.T) {
 	if config.SessionTTLMin != 5*time.Second {
 		t.Fatalf("bad: %s %#v", config.SessionTTLMin.String(), config)
 	}
+
+	// DisableReap
+	input = `{"disable_reap": true}`
+	config, err = DecodeConfig(bytes.NewReader([]byte(input)))
+	if err != nil {
+		t.Fatalf("err: %s", err)
+	}
+
+	if config.DisableReap != true {
+		t.Fatalf("bad: reap not disabled: %#v", config)
+	}
 }
 
 func TestDecodeConfig_invalidKeys(t *testing.T) {
@@ -1157,6 +1168,7 @@ func TestMergeConfig(t *testing.T) {
 		CheckUpdateIntervalRaw: "8m",
 		RetryIntervalRaw:       "10s",
 		RetryIntervalWanRaw:    "10s",
+		DisableReap:            false,
 	}
 
 	b := &Config{
@@ -1266,6 +1278,7 @@ func TestMergeConfig(t *testing.T) {
 			RPC:        &net.TCPAddr{},
 			RPCRaw:     "127.0.0.5:1233",
 		},
+		DisableReap: true,
 	}
 
 	c := MergeConfig(a, b)
diff --git a/website/source/docs/agent/options.html.markdown b/website/source/docs/agent/options.html.markdown
index cadd50621..aac394f32 100644
--- a/website/source/docs/agent/options.html.markdown
+++ b/website/source/docs/agent/options.html.markdown
@@ -407,6 +407,10 @@ definitions support being updated during a reload.
   `disable_anonymous_signature`</a> Disables providing an anonymous signature for de-duplication
   with the update check. See [`disable_update_check`](#disable_update_check).
 
+* <a name="disable_reap"></a><a href="#disable_reap">
+  `disable_reap`</a> will prevent Consul from automatically reaping child processes if it
+  detects it is running as PID 1, such as in a Docker container.
+
 * <a name="disable_remote_exec"></a><a href="#disable_remote_exec">`disable_remote_exec`</a>
   Disables support for remote execution. When set to true, the agent will ignore any incoming
   remote exec requests.