Merge pull request #51533 from shyamjvs/master-creation-retry

Automatic merge from submit-queue Retry master instance creation in case of retriable error (with sleep) To help with our 5k-node CI tests failing to startup the cluster. And also towards the greater goal - https://github.com/kubernetes/kubernetes/issues/43140 cc @kubernetes/sig-scalability-misc @kubernetes/sig-cluster-lifecycle-misc

Merge pull request #51533 from shyamjvs/master-creation-retry
e3210c6c · Kubernetes Submit Queue · GitHub · 0596891e · 2595f664 · e3210c6c
Commit e3210c6c authored Aug 30, 2017 by Kubernetes Submit Queue Committed by GitHub Aug 30, 2017
Hide whitespace changes
Inline Side-by-side

Showing with 58 additions and 26 deletions

master-helper.sh cluster/gce/container-linux/master-helper.sh +29 -13

master-helper.sh cluster/gce/gci/master-helper.sh +29 -13

No files found.
--- a/cluster/gce/container-linux/master-helper.sh
+++ b/cluster/gce/container-linux/master-helper.sh
@@ -66,6 +66,7 @@ function replicate-master-instance() {

 function create-master-instance-internal() {
  local gcloud="gcloud"
+  local retries=5
  if [[ "${ENABLE_IP_ALIASES:-}" == 'true' ]]; then
    gcloud="gcloud beta"
  fi
@@ -93,19 +94,34 @@ function create-master-instance-internal() {
  disk="${disk},boot=no"
  disk="${disk},auto-delete=no"

-  ${gcloud} compute instances create "${master_name}" \
-    --project "${PROJECT}" \
-    --zone "${ZONE}" \
-    --machine-type "${MASTER_SIZE}" \
-    --image-project="${MASTER_IMAGE_PROJECT}" \
-    --image "${MASTER_IMAGE}" \
-    --tags "${MASTER_TAG}" \
-    --scopes "storage-ro,compute-rw,monitoring,logging-write" \
-    --metadata-from-file "${metadata}" \
-    --disk "${disk}" \
-    --boot-disk-size "${MASTER_ROOT_DISK_SIZE:-30}" \
-    ${preemptible_master} \
-    ${network}
+  for attempt in $(seq 1 ${retries}); do
+    if result=$(${gcloud} compute instances create "${master_name}" \
+      --project "${PROJECT}" \
+      --zone "${ZONE}" \
+      --machine-type "${MASTER_SIZE}" \
+      --image-project="${MASTER_IMAGE_PROJECT}" \
+      --image "${MASTER_IMAGE}" \
+      --tags "${MASTER_TAG}" \
+      --scopes "storage-ro,compute-rw,monitoring,logging-write" \
+      --metadata-from-file "${metadata}" \
+      --disk "${disk}" \
+      --boot-disk-size "${MASTER_ROOT_DISK_SIZE}" \
+      ${preemptible_master} \
+      ${network} 2>&1); then
+      echo "${result}" >&2
+      return 0
+    else
+      echo "${result}" >&2
+      if [[ ! "${result}" =~ "try again later" ]]; then
+        echo "Failed to create master instance due to non-retryable error" >&2
+        return 1
+      fi
+      sleep 10
+    fi
+  done
+
+  echo "Failed to create master instance despite ${retries} attempts" >&2
+  return 1
 }

 function get-metadata() {

--- a/cluster/gce/gci/master-helper.sh
+++ b/cluster/gce/gci/master-helper.sh
@@ -75,6 +75,7 @@ function replicate-master-instance() {

 function create-master-instance-internal() {
  local gcloud="gcloud"
+  local retries=5
  if [[ "${ENABLE_IP_ALIASES:-}" == 'true' ]]; then
    gcloud="gcloud beta"
  fi
@@ -106,19 +107,34 @@ function create-master-instance-internal() {
  disk="${disk},boot=no"
  disk="${disk},auto-delete=no"

-  ${gcloud} compute instances create "${master_name}" \
-    --project "${PROJECT}" \
-    --zone "${ZONE}" \
-    --machine-type "${MASTER_SIZE}" \
-    --image-project="${MASTER_IMAGE_PROJECT}" \
-    --image "${MASTER_IMAGE}" \
-    --tags "${MASTER_TAG}" \
-    --scopes "storage-ro,compute-rw,monitoring,logging-write" \
-    --metadata-from-file "${metadata}" \
-    --disk "${disk}" \
-    --boot-disk-size "${MASTER_ROOT_DISK_SIZE}" \
-    ${preemptible_master} \
-    ${network}
+  for attempt in $(seq 1 ${retries}); do
+    if result=$(${gcloud} compute instances create "${master_name}" \
+      --project "${PROJECT}" \
+      --zone "${ZONE}" \
+      --machine-type "${MASTER_SIZE}" \
+      --image-project="${MASTER_IMAGE_PROJECT}" \
+      --image "${MASTER_IMAGE}" \
+      --tags "${MASTER_TAG}" \
+      --scopes "storage-ro,compute-rw,monitoring,logging-write" \
+      --metadata-from-file "${metadata}" \
+      --disk "${disk}" \
+      --boot-disk-size "${MASTER_ROOT_DISK_SIZE}" \
+      ${preemptible_master} \
+      ${network} 2>&1); then
+      echo "${result}" >&2
+      return 0
+    else
+      echo "${result}" >&2
+      if [[ ! "${result}" =~ "try again later" ]]; then
+        echo "Failed to create master instance due to non-retryable error" >&2
+        return 1
+      fi
+      sleep 10
+    fi
+  done
+
+  echo "Failed to create master instance despite ${retries} attempts" >&2
+  return 1
 }

 function get-metadata() {