From 1e791aef776d2364ed55b0e5bdc700535171a6e4 Mon Sep 17 00:00:00 2001
From: Derek McGowan <derek@mcgstyle.net>
Date: Tue, 4 Sep 2018 11:00:28 -0700
Subject: [PATCH] Fix supervisor healthcheck throttling

Fix default case causing the throttling to not be used.
Ensure that nil client condition is handled.

Signed-off-by: Derek McGowan <derek@mcgstyle.net>
(cherry picked from commit c3e32938430e03a316311f9e4fbdb743e492a07e)
Signed-off-by: Andrew Hsu <andrewhsu@docker.com>
Upstream-commit: f121eccf29576ce5d4b8256a71a9d32ee688ff7d
Component: engine
---
 .../libcontainerd/supervisor/remote_daemon.go | 61 +++++++++++--------
 1 file changed, 35 insertions(+), 26 deletions(-)

diff --git a/components/engine/libcontainerd/supervisor/remote_daemon.go b/components/engine/libcontainerd/supervisor/remote_daemon.go
index b520d48671..182984aeca 100644
--- a/components/engine/libcontainerd/supervisor/remote_daemon.go
+++ b/components/engine/libcontainerd/supervisor/remote_daemon.go
@@ -245,20 +245,26 @@ func (r *remote) monitorDaemon(ctx context.Context) {
 	}()
 
 	for {
-		select {
-		case <-ctx.Done():
-			r.logger.Info("stopping healthcheck following graceful shutdown")
-			if client != nil {
-				client.Close()
+		if delay != nil {
+			select {
+			case <-ctx.Done():
+				r.logger.Info("stopping healthcheck following graceful shutdown")
+				if client != nil {
+					client.Close()
+				}
+				return
+			case <-delay:
 			}
-			return
-		case <-delay:
-		default:
 		}
 
 		if r.daemonPid == -1 {
 			if r.daemonWaitCh != nil {
-				<-r.daemonWaitCh
+				select {
+				case <-ctx.Done():
+					r.logger.Info("stopping containerd startup following graceful shutdown")
+					return
+				case <-r.daemonWaitCh:
+				}
 			}
 
 			os.RemoveAll(r.GRPC.Address)
@@ -276,26 +282,28 @@ func (r *remote) monitorDaemon(ctx context.Context) {
 			}
 		}
 
-		tctx, cancel := context.WithTimeout(ctx, healthCheckTimeout)
-		_, err := client.IsServing(tctx)
-		cancel()
-		if err == nil {
-			if !started {
-				close(r.daemonStartCh)
-				started = true
+		if client != nil {
+			tctx, cancel := context.WithTimeout(ctx, healthCheckTimeout)
+			_, err := client.IsServing(tctx)
+			cancel()
+			if err == nil {
+				if !started {
+					close(r.daemonStartCh)
+					started = true
+				}
+
+				transientFailureCount = 0
+				delay = time.After(500 * time.Millisecond)
+				continue
 			}
 
-			transientFailureCount = 0
-			delay = time.After(500 * time.Millisecond)
-			continue
-		}
+			r.logger.WithError(err).WithField("binary", binaryName).Debug("daemon is not responding")
 
-		r.logger.WithError(err).WithField("binary", binaryName).Debug("daemon is not responding")
-
-		transientFailureCount++
-		if transientFailureCount < maxConnectionRetryCount || system.IsProcessAlive(r.daemonPid) {
-			delay = time.After(time.Duration(transientFailureCount) * 200 * time.Millisecond)
-			continue
+			transientFailureCount++
+			if transientFailureCount < maxConnectionRetryCount || system.IsProcessAlive(r.daemonPid) {
+				delay = time.After(time.Duration(transientFailureCount) * 200 * time.Millisecond)
+				continue
+			}
 		}
 
 		if system.IsProcessAlive(r.daemonPid) {
@@ -304,6 +312,7 @@ func (r *remote) monitorDaemon(ctx context.Context) {
 		}
 
 		client.Close()
+		client = nil
 		r.daemonPid = -1
 		delay = nil
 		transientFailureCount = 0