From 424685321191af4f3b9c3fe88581f454448718a4 Mon Sep 17 00:00:00 2001
From: Random-Liu <taotaotheripper@gmail.com>
Date: Fri, 8 Jul 2016 10:45:20 -0700
Subject: [PATCH] Prevent kube-proxy from panicing when sysfs is mounted as
 read-only. Send a node event when this happens and hint to the administrator
 about the remediation.

---
 cmd/kube-proxy/app/conntrack.go | 40 +++++++++++++++++++++++++++++++++
 cmd/kube-proxy/app/server.go    | 15 +++++++++++--
 2 files changed, 53 insertions(+), 2 deletions(-)

diff --git a/cmd/kube-proxy/app/conntrack.go b/cmd/kube-proxy/app/conntrack.go
index b7323033e15..001d9911193 100644
--- a/cmd/kube-proxy/app/conntrack.go
+++ b/cmd/kube-proxy/app/conntrack.go
@@ -17,11 +17,13 @@ limitations under the License.
 package app
 
 import (
+	"errors"
 	"io/ioutil"
 	"strconv"
 
 	"github.com/golang/glog"
 
+	"k8s.io/kubernetes/pkg/util/mount"
 	"k8s.io/kubernetes/pkg/util/sysctl"
 )
 
@@ -32,11 +34,25 @@ type Conntracker interface {
 
 type realConntracker struct{}
 
+var readOnlySysFSError = errors.New("ReadOnlySysFS")
+
 func (realConntracker) SetMax(max int) error {
 	glog.Infof("Setting nf_conntrack_max to %d", max)
 	if err := sysctl.SetSysctl("net/netfilter/nf_conntrack_max", max); err != nil {
 		return err
 	}
+	// sysfs is expected to be mounted as 'rw'. However, it may be unexpectedly mounted as
+	// 'ro' by docker because of a known docker issue (https://github.com/docker/docker/issues/24000).
+	// Setting conntrack will fail when sysfs is readonly. When that happens, we don't set conntrack
+	// hashsize and return a special error readOnlySysFSError here. The caller should deal with
+	// readOnlySysFSError differently.
+	writable, err := isSysFSWritable()
+	if err != nil {
+		return err
+	}
+	if !writable {
+		return readOnlySysFSError
+	}
 	// TODO: generify this and sysctl to a new sysfs.WriteInt()
 	glog.Infof("Setting conntrack hashsize to %d", max/4)
 	return ioutil.WriteFile("/sys/module/nf_conntrack/parameters/hashsize", []byte(strconv.Itoa(max/4)), 0640)
@@ -46,3 +62,27 @@ func (realConntracker) SetTCPEstablishedTimeout(seconds int) error {
 	glog.Infof("Setting nf_conntrack_tcp_timeout_established to %d", seconds)
 	return sysctl.SetSysctl("net/netfilter/nf_conntrack_tcp_timeout_established", seconds)
 }
+
+// isSysFSWritable checks /proc/mounts to see whether sysfs is 'rw' or not.
+func isSysFSWritable() (bool, error) {
+	const permWritable = "rw"
+	const sysfsDevice = "sysfs"
+	m := mount.New()
+	mountPoints, err := m.List()
+	if err != nil {
+		glog.Errorf("failed to list mount points: %v", err)
+		return false, err
+	}
+	for _, mountPoint := range mountPoints {
+		if mountPoint.Device != sysfsDevice {
+			continue
+		}
+		// Check whether sysfs is 'rw'
+		if len(mountPoint.Opts) > 0 && mountPoint.Opts[0] == permWritable {
+			return true, nil
+		}
+		glog.Errorf("sysfs is not writable: %+v", mountPoint)
+		break
+	}
+	return false, nil
+}
diff --git a/cmd/kube-proxy/app/server.go b/cmd/kube-proxy/app/server.go
index dc5d4df0007..51e8d2eec3f 100644
--- a/cmd/kube-proxy/app/server.go
+++ b/cmd/kube-proxy/app/server.go
@@ -299,8 +299,19 @@ func (s *ProxyServer) Run() error {
 	// Tune conntrack, if requested
 	if s.Conntracker != nil {
 		if s.Config.ConntrackMax > 0 {
-			if err := s.Conntracker.SetMax(int(s.Config.ConntrackMax)); err != nil {
-				return err
+			err := s.Conntracker.SetMax(int(s.Config.ConntrackMax))
+			if err != nil {
+				if err != readOnlySysFSError {
+					return err
+				}
+				// readOnlySysFSError is caused by a known docker issue (https://github.com/docker/docker/issues/24000),
+				// the only remediation we know is to restart the docker daemon.
+				// Here we'll send an node event with specific reason and message, the
+				// administrator should decide whether and how to handle this issue,
+				// whether to drain the node and restart docker.
+				// TODO(random-liu): Remove this when the docker bug is fixed.
+				const message = "DOCKER RESTART NEEDED (docker issue #24000): /sys is read-only: can't raise conntrack limits, problems may arise later."
+				s.Recorder.Eventf(s.Config.NodeRef, api.EventTypeWarning, err.Error(), message)
 			}
 		}
 		if s.Config.ConntrackTCPEstablishedTimeout.Duration > 0 {