From 78dabbdb7f5813d257d054fd8e3d57903e5034ff Mon Sep 17 00:00:00 2001
From: Brendan Burns <bburns@google.com>
Date: Tue, 21 Apr 2015 15:27:38 -0700
Subject: [PATCH] Fix the ssh-to-node to actually fail on failures.

---
 cluster/gce/util.sh |  6 +++++-
 cluster/gke/util.sh | 11 +++++++++--
 2 files changed, 14 insertions(+), 3 deletions(-)

diff --git a/cluster/gce/util.sh b/cluster/gce/util.sh
index c5f3fce8f27..c2267902cce 100755
--- a/cluster/gce/util.sh
+++ b/cluster/gce/util.sh
@@ -967,11 +967,15 @@ function test-teardown {
 function ssh-to-node {
   local node="$1"
   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 "${cmd}"; then
+    if gcloud compute ssh --ssh-flag="-o LogLevel=quiet" --project "${PROJECT}" --zone="${ZONE}" "${node}" --command "echo test"; 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}"
 }
 
 # Restart the kube-proxy on a node ($1)
diff --git a/cluster/gke/util.sh b/cluster/gke/util.sh
index c4451adad26..ea66997dac8 100755
--- a/cluster/gke/util.sh
+++ b/cluster/gke/util.sh
@@ -238,8 +238,15 @@ function ssh-to-node() {
 
   local node="$1"
   local cmd="$2"
-  "${GCLOUD}" compute ssh --ssh-flag="-o LogLevel=quiet" --project "${PROJECT}" \
-    --zone="${ZONE}" "${node}" --command "${cmd}"
+  # 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"; 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}"
 }
 
 # Restart the kube-proxy on a node ($1)