Fix up e2e tests.

cluster/gce/config-default.sh

@@ -15,6 +15,8 @@
 # limitations under the License.
 
 # TODO(jbeda): Provide a way to override project
+# gcloud multiplexing for shared GCE/GKE tests.
+GCLOUD=gcloud
 ZONE=us-central1-b
 MASTER_SIZE=n1-standard-1
 MINION_SIZE=n1-standard-1

cluster/gce/config-test.sh

@@ -15,6 +15,8 @@
 # limitations under the License.
 
 # TODO(jbeda): Provide a way to override project
+# gcloud multiplexing for shared GCE/GKE tests.
+GCLOUD=gcloud
 ZONE=us-central1-b
 MASTER_SIZE=g1-small
 MINION_SIZE=g1-small

hack/e2e-suite/basic.sh

@@ -25,10 +25,6 @@ KUBE_ROOT=$(dirname "${BASH_SOURCE}")/../..
 source "${KUBE_ROOT}/cluster/kube-env.sh"
 source "${KUBE_ROOT}/cluster/$KUBERNETES_PROVIDER/util.sh"
 
-# Launch some pods.
-num_pods=2
-$KUBECFG -p 8080:9376 run kubernetes/serve_hostname ${num_pods} my-hostname
-
 function teardown() {
   echo "Cleaning up test artifacts"
   $KUBECFG stop my-hostname
@@ -37,7 +33,21 @@ function teardown() {
 
 trap "teardown" EXIT
 
-pod_id_list=$($KUBECFG '-template={{range.items}}{{.id}} {{end}}' -l replicationController=my-hostname list pods)
+# Determine which pod image to launch (e.g. private.sh launches a different one).
+pod_img_srv="${POD_IMG_SRV:-kubernetes/serve_hostname}"
+
+# Launch some pods.
+num_pods=2
+$KUBECFG -p 8080:9376 run "${pod_img_srv}" ${num_pods} my-hostname
+
+# List the pods.
+pod_id_list=$($KUBECFG '-template={{range.items}}{{.id}} {{end}}' -l name=my-hostname list pods)
+echo "pod_id_list: ${pod_id_list}"
+if [[ -z "${pod_id_list:-}" ]]; then
+  echo "Pod ID list is empty. It should have a set of pods to verify."
+  exit 1
+fi
+
 # Pod turn up on a clean cluster can take a while for the docker image pull.
 all_running=0
 for i in $(seq 1 24); do
@@ -45,7 +55,7 @@ for i in $(seq 1 24); do
   sleep 5
   all_running=1
   for id in $pod_id_list; do
-    current_status=$($KUBECFG -template '{{.currentState.status}}' get pods/$id) || true
+    current_status=$($KUBECFG '-template={{.currentState.status}}' get pods/$id) || true
     if [[ "$current_status" != "Running" ]]; then
       all_running=0
       break
@@ -60,22 +70,28 @@ if [[ "${all_running}" == 0 ]]; then
   exit 1
 fi
 
-# Get minion IP addresses
-detect-minions
-
 # let images stabilize
 echo "Letting images stabilize"
 sleep 5
 
 # Verify that something is listening.
 for id in ${pod_id_list}; do
-  ip=$($KUBECFG -template '{{.currentState.hostIP}}' get pods/$id)
+  ip=$($KUBECFG '-template={{.currentState.hostIP}}' get pods/$id)
   echo "Trying to reach server that should be running at ${ip}:8080..."
-  ok=0
+  server_running=0
   for i in $(seq 1 5); do
-    curl --connect-timeout 1 "http://${ip}:8080" >/dev/null 2>&1 && ok=1 && break
+    echo "--- trial ${i}"
+    output=$(curl -s -connect-timeout 1 "http://${ip}:8080" || true)
+    if echo $output | grep "${id}" &> /dev/null; then
+      server_running=1
+      break
+    fi
     sleep 2
   done
+  if [[ "${server_running}" -ne 1 ]]; then
+    echo "Server never running at ${ip}:8080..."
+    exit 1
+  fi
 done
 
 exit 0

hack/e2e-suite/certs.sh

@@ -25,14 +25,18 @@ KUBE_ROOT=$(dirname "${BASH_SOURCE}")/../..
 source "${KUBE_ROOT}/cluster/kube-env.sh"
 source "${KUBE_ROOT}/cluster/${KUBERNETES_PROVIDER}/util.sh"
 
-if [[ "${KUBERNETES_PROVIDER}" != "gce" ]]; then
-  echo "Skipping certs test on non-gce provider."
+if [[ "${KUBERNETES_PROVIDER}" != "gce" ]] && [[ "${KUBERNETES_PROVIDER}" != "gke" ]]; then
+  echo "WARNING: Skipping certs.sh for cloud provider: ${KUBERNETES_PROVIDER}."
   exit 0
 fi
 
+# Set KUBE_MASTER
+detect-master
+
 # IMPORTANT: there are upstream things that rely on these files.
 # Do *not* fix this test by changing this path, unless you _really_ know
 # what you are doing.
 for file in kubecfg.key kubecfg.crt ca.crt; do
-  gcloud compute ssh --zone="${ZONE}" "${MASTER_NAME}" --command "ls /srv/kubernetes/${file}"
+  echo "Checking for ${file}"
+  "${GCLOUD}" compute ssh --zone="${ZONE}" "${KUBE_MASTER}" --command "ls /srv/kubernetes/${file}"
 done

hack/e2e-suite/monitoring.sh

@@ -26,22 +26,29 @@ KUBE_ROOT=$(dirname "${BASH_SOURCE}")/../..
 source "${KUBE_ROOT}/cluster/kube-env.sh"
 source "${KUBE_ROOT}/cluster/$KUBERNETES_PROVIDER/util.sh"
 
+if [[ "${KUBERNETES_PROVIDER}" != "gce" ]] && [[ "${KUBERNETES_PROVIDER}" != "gke" ]]; then
+  echo "WARNING: Skipping monitoring.sh for cloud provider: ${KUBERNETES_PROVIDER}."
+  exit 0
+fi
+
 MONITORING="${KUBE_ROOT}/examples/monitoring"
 KUBECTL="${KUBE_ROOT}/cluster/kubectl.sh"
 MONITORING_FIREWALL_RULE="monitoring-test"
 
 function setup {
   detect-project
-  if ! gcloud compute firewall-rules describe $MONITORING_FIREWALL_RULE &>/dev/null; then
-    if ! gcloud compute firewall-rules create $MONITORING_FIREWALL_RULE \
+
+  if ! "${GCLOUD}" compute firewall-rules describe $MONITORING_FIREWALL_RULE &> /dev/null; then
+    if ! "${GCLOUD}" compute firewall-rules create $MONITORING_FIREWALL_RULE \
       --project "${PROJECT}" \
-      --network "e2e" \
+      --network "${NETWORK}" \
       --quiet \
       --allow tcp:80 tcp:8083 tcp:8086 tcp:9200; then
       echo "Failed to set up firewall for monitoring" && false
     fi
   fi
-  "${KUBECTL}" create -f "${MONITORING}/influx-grafana-pod.json" 
+
+  "${KUBECTL}" create -f "${MONITORING}/influx-grafana-pod.json"
   "${KUBECTL}" create -f "${MONITORING}/influx-grafana-service.json"
   "${KUBECTL}" create -f "${MONITORING}/heapster-pod.json"
 }
@@ -51,8 +58,8 @@ function cleanup {
   "${KUBECTL}" delete -f "${MONITORING}/influx-grafana-pod.json" || true
   "${KUBECTL}" delete -f "${MONITORING}/influx-grafana-service.json" || true
   "${KUBECTL}" delete -f "${MONITORING}/heapster-pod.json" || true
-  if gcloud compute firewall-rules describe $MONITORING_FIREWALL_RULE &> /dev/null; then
-    gcloud compute firewall-rules delete \
+  if "${GCLOUD}" compute firewall-rules describe $MONITORING_FIREWALL_RULE &> /dev/null; then
+    "${GCLOUD}" compute firewall-rules delete \
       --project "${PROJECT}" \
       --quiet \
       $MONITORING_FIREWALL_RULE || true
@@ -71,7 +78,7 @@ function influx-data-exists {
 
 function wait-for-pods {
   local running=false
-  for i in `seq 1 20`; do   
+  for i in `seq 1 20`; do
     sleep 20
     if "${KUBECTL}" get pods influx-grafana | grep Running &> /dev/null \
       && "${KUBECTL}" get pods heapster | grep Running &> /dev/null; then

hack/e2e-suite/pd.sh

@@ -25,9 +25,9 @@ KUBE_ROOT=$(dirname "${BASH_SOURCE}")/../..
 source "${KUBE_ROOT}/cluster/kube-env.sh"
 source "${KUBE_ROOT}/cluster/$KUBERNETES_PROVIDER/util.sh"
 
-if [[ "$KUBERNETES_PROVIDER" != "gce" ]]; then
-    echo "PD test is only run for GCE"
-    return 0
+if [[ "$KUBERNETES_PROVIDER" != "gce" ]] && [[ "$KUBERNETES_PROVIDER" != "gke" ]]; then
+  echo "WARNING: Skipping pd.sh for cloud provider: ${KUBERNETES_PROVIDER}."
+  exit 0
 fi
 
 disk_name="e2e-$(date +%H-%M-%s)"

hack/e2e-suite/private.sh

@@ -21,66 +21,16 @@ set -o errexit
 set -o nounset
 set -o pipefail
 
-if [[ "${KUBERNETES_PROVIDER:-gce}" != "gce" ]]; then
-  echo WARNING: Skipping private.sh for cloud provider: $KUBERNETES_PROVIDER.
-  exit 0
-fi
-
 KUBE_ROOT=$(dirname "${BASH_SOURCE}")/../..
 source "${KUBE_ROOT}/cluster/kube-env.sh"
 source "${KUBE_ROOT}/cluster/$KUBERNETES_PROVIDER/util.sh"
 
-# Launch some pods.
-num_pods=2
-$KUBECFG -p 8080:9376 run container.cloud.google.com/_b_k8s_test/serve_hostname ${num_pods} my-hostname
-
-function teardown() {
-  echo "Cleaning up test artifacts"
-  $KUBECFG stop my-hostname
-  $KUBECFG rm my-hostname
-}
-
-trap "teardown" EXIT
-
-pod_id_list=$($KUBECFG '-template={{range.items}}{{.id}} {{end}}' -l replicationController=my-hostname list pods)
-# Pod turn up on a clean cluster can take a while for the docker image pull.
-all_running=0
-for i in $(seq 1 24); do
-  echo "Waiting for pods to come up."
-  sleep 5
-  all_running=1
-  for id in $pod_id_list; do
-    current_status=$($KUBECFG -template '{{.currentState.status}}' get pods/$id) || true
-    if [[ "$current_status" != "Running" ]]; then
-      all_running=0
-      break
-    fi
-  done
-  if [[ "${all_running}" == 1 ]]; then
-    break
-  fi
-done
-if [[ "${all_running}" == 0 ]]; then
-  echo "Pods did not come up in time"
-  exit 1
+# Private image works only on GCE and GKE.
+if [[ "${KUBERNETES_PROVIDER}" != "gce" ]] && [[ "${KUBERNETES_PROVIDER}" != "gke" ]]; then
+  echo "WARNING: Skipping private.sh for cloud provider: ${KUBERNETES_PROVIDER}."
+  exit 0
 fi
 
-# Get minion IP addresses
-detect-minions
-
-# let images stabilize
-echo "Letting images stabilize"
-sleep 5
-
-# Verify that something is listening.
-for id in ${pod_id_list}; do
-  ip=$($KUBECFG -template '{{.currentState.hostIP}}' get pods/$id)
-  echo "Trying to reach server that should be running at ${ip}:8080..."
-  ok=0
-  for i in $(seq 1 5); do
-    curl --connect-timeout 1 "http://${ip}:8080" >/dev/null 2>&1 && ok=1 && break
-    sleep 2
-  done
-done
-
-exit 0
+# Run the basic.sh test, but using this image.
+export POD_IMG_SRV="container.cloud.google.com/_b_k8s_test/serve_hostname"
+source "${KUBE_ROOT}/hack/e2e-suite/basic.sh"

hack/e2e-suite/services.sh

@@ -200,6 +200,7 @@ function wait_for_pods() {
 #   $5: pod IDs
 function wait_for_service_up() {
   local i
+  local found_pods
   for i in $(seq 1 20); do
     results=($(ssh-to-node "${test_node}" "
         set -e;
@@ -207,7 +208,9 @@ function wait_for_service_up() {
           curl -s --connect-timeout 1 http://$2:$3;
         done | sort | uniq
         "))
+
     found_pods=$(sort_args "${results[@]:+${results[@]}}")
+    echo "Checking if ${found_pods} == ${5}"
     if [[ "${found_pods}" == "$5" ]]; then
       break
     fi
@@ -299,6 +302,7 @@ fi
 #
 # Test 1: Prove that the service portal is alive.
 #
+echo "Test 1: Prove that the service portal is alive."
 echo "Verifying the portals from the host"
 wait_for_service_up "${svc1_name}" "${svc1_ip}" "${svc1_port}" \
     "${svc1_count}" "${svc1_pods}"
@@ -321,6 +325,7 @@ verify_from_container "${svc2_name}" "${svc2_ip}" "${svc2_port}" \
 #
 # Test 2: Bounce the proxy and make sure the portal comes back.
 #
+echo "Test 2: Bounce the proxy and make sure the portal comes back."
 echo "Restarting kube-proxy"
 restart-kube-proxy "${test_node}"
 echo "Verifying the portals from the host"
@@ -337,6 +342,7 @@ verify_from_container "${svc2_name}" "${svc2_ip}" "${svc2_port}" \
 #
 # Test 3: Stop one service and make sure it is gone.
 #
+echo "Test 3: Stop one service and make sure it is gone."
 stop_service "${svc1_name}"
 wait_for_service_down "${svc1_name}" "${svc1_ip}" "${svc1_port}"
 
@@ -344,6 +350,7 @@ wait_for_service_down "${svc1_name}" "${svc1_ip}" "${svc1_port}"
 # Test 4: Bring up another service.
 # TODO: Actually add a test to force re-use.
 #
+echo "Test 4: Bring up another service."
 svc3_name="service3"
 svc3_port=80
 svc3_count=3
@@ -369,6 +376,7 @@ verify_from_container "${svc3_name}" "${svc3_ip}" "${svc3_port}" \
 #
 # Test 5: Remove the iptables rules, make sure they come back.
 #
+echo "Test 5: Remove the iptables rules, make sure they come back."
 echo "Manually removing iptables rules"
 ssh-to-node "${test_node}" "sudo iptables -t nat -F KUBE-PROXY"
 echo "Verifying the portals from the host"
@@ -381,6 +389,7 @@ verify_from_container "${svc3_name}" "${svc3_ip}" "${svc3_port}" \
 #
 # Test 6: Restart the master, make sure portals come back.
 #
+echo "Test 6: Restart the master, make sure portals come back."
 echo "Restarting the master"
 ssh-to-node "${master}" "sudo /etc/init.d/kube-apiserver restart"
 sleep 5
@@ -394,6 +403,7 @@ verify_from_container "${svc3_name}" "${svc3_ip}" "${svc3_port}" \
 #
 # Test 7: Bring up another service, make sure it does not re-use Portal IPs.
 #
+echo "Test 7: Bring up another service, make sure it does not re-use Portal IPs."
 svc4_name="service4"
 svc4_port=80
 svc4_count=3

hack/e2e.go

@@ -165,7 +165,7 @@ func Up() bool {
 
 // Is the e2e cluster up?
 func IsUp() bool {
-	return runBash("get status", `$KUBECFG -server_version`)
+	return runBash("get status", `$KUBECTL version`)
 }
 
 func tryUp() bool {