From e660adbad83e3a793f6cae628402f7ada817c016 Mon Sep 17 00:00:00 2001
From: Joe Finney <spxtr@google.com>
Date: Mon, 7 Mar 2016 13:29:04 -0800
Subject: [PATCH] Add connect timeout to ssh-to-node.

---
 cluster/aws/util.sh            | 4 +++-
 cluster/gce/util.sh            | 4 ++--
 cluster/gke/util.sh            | 4 ++--
 cluster/libvirt-coreos/util.sh | 2 +-
 4 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/cluster/aws/util.sh b/cluster/aws/util.sh
index 96882dbd59c..13639fa56fb 100755
--- a/cluster/aws/util.sh
+++ b/cluster/aws/util.sh
@@ -1511,10 +1511,12 @@ function ssh-to-node {
   fi
 
   for try in $(seq 1 5); do
-    if ssh -oLogLevel=quiet -oStrictHostKeyChecking=no -i "${AWS_SSH_KEY}" ${SSH_USER}@${ip} "${cmd}"; then
+    if ssh -oLogLevel=quiet -oConnectTimeout=30 -oStrictHostKeyChecking=no -i "${AWS_SSH_KEY}" ${SSH_USER}@${ip} "echo test > /dev/null"; then
       break
     fi
+    sleep 5
   done
+  ssh -oLogLevel=quiet -oConnectTimeout=30 -oStrictHostKeyChecking=no -i "${AWS_SSH_KEY}" ${SSH_USER}@${ip} "${cmd}"
 }
 
 # Restart the kube-proxy on a node ($1)
diff --git a/cluster/gce/util.sh b/cluster/gce/util.sh
index 724834d9d03..78c8a7aba55 100755
--- a/cluster/gce/util.sh
+++ b/cluster/gce/util.sh
@@ -1319,13 +1319,13 @@ function ssh-to-node {
   local cmd="$2"
   # Loop until we can successfully ssh into the box
   for try in $(seq 1 5); do
-    if gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "echo test > /dev/null"; then
+    if gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --ssh-flag="-o ConnectTimeout=30" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "echo test > /dev/null"; then
       break
     fi
     sleep 5
   done
   # Then actually try the command.
-  gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "${cmd}"
+  gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --ssh-flag="-o ConnectTimeout=30" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "${cmd}"
 }
 
 # Restart the kube-proxy on a node ($1)
diff --git a/cluster/gke/util.sh b/cluster/gke/util.sh
index cc1931bffb7..e2a6e112965 100755
--- a/cluster/gke/util.sh
+++ b/cluster/gke/util.sh
@@ -263,13 +263,13 @@ function ssh-to-node() {
   local cmd="$2"
   # Loop until we can successfully ssh into the box
   for try in $(seq 1 5); do
-    if gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "echo test > /dev/null"; then
+    if gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --ssh-flag="-o ConnectTimeout=30" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "echo test > /dev/null"; then
       break
     fi
     sleep 5
   done
   # Then actually try the command.
-  gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "${cmd}"
+  gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --ssh-flag="-o ConnectTimeout=30" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "${cmd}"
 }
 
 # Restart the kube-proxy on a node ($1)
diff --git a/cluster/libvirt-coreos/util.sh b/cluster/libvirt-coreos/util.sh
index f70d202e769..f10ccfddb90 100644
--- a/cluster/libvirt-coreos/util.sh
+++ b/cluster/libvirt-coreos/util.sh
@@ -345,7 +345,7 @@ function ssh-to-node {
   if [[ -z "$machine" ]]; then
       echo "$node is an unknown machine to ssh to" >&2
   fi
-  ssh -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o ControlMaster=no "core@$machine" "$cmd"
+  ssh -o ConnectTimeout=30 -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o ControlMaster=no "core@$machine" "$cmd"
 }
 
 # Restart the kube-proxy on a node ($1)