Adjust sample Docker and Kubelet configuration

Docker:
* max-concurrent-downloads - This is a minor optimization for testing.
* oom-score-adjust - This make the kernel less likely to kill the docker
  daemon when it runs out of memory.

Kubelet:
* serialize-image-pulls=false - Combined with the above
  max-concurrent-downloads, improves local resiliency gate runtime from
  ~25 mins to ~23.  Nearly 10% off the testing cycle for a config
  change :)

Change-Id: I9dacbeee9615d5f6fd2577e2af4842e3603a11cc

examples/basic/Docker.yaml

@@ -11,5 +11,7 @@ data:
     insecure-registries:
       - registry:5000
     live-restore: true
+    max-concurrent-downloads: 10
+    oom-score-adjust: -999
     storage-driver: overlay2
 ...

examples/basic/Kubelet.yaml

@@ -13,6 +13,7 @@ data:
     - --eviction-max-pod-grace-period=-1
     - --network-plugin=cni
     - --node-status-update-frequency=5s
+    - --serialize-image-pulls=false
     - --v=5
   images:
     pause: gcr.io/google_containers/pause-amd64:3.0

examples/complete/Docker.yaml

@@ -12,5 +12,7 @@ data:
     insecure-registries:
       - registry:5000
     live-restore: true
+    max-concurrent-downloads: 10
+    oom-score-adjust: -999
     storage-driver: overlay2
 ...

examples/complete/Kubelet.yaml

@@ -14,6 +14,7 @@ data:
     - --eviction-max-pod-grace-period=-1
     - --network-plugin=cni
     - --node-status-update-frequency=5s
+    - --serialize-image-pulls=false
     - --v=5
   images:
     pause: gcr.io/google_containers/pause-amd64:3.0

tools/g2/stages/join-nodes.sh

@@ -90,13 +90,22 @@ for NAME in "${NODES[@]}"; do
     fi
 
     log "Checking Promenade API health"
-    ssh_cmd "${VIA}" curl -v "${CURL_ARGS[@]}" \
-        "${BASE_PROM_URL}/api/v1.0/health"
-    log "Promenade API healthy"
+    MAX_HEALTH_ATTEMPTS=6
+    for attempt in $(seq ${MAX_HEALTH_ATTEMPTS}); do
+        if ssh_cmd "${VIA}" curl -v "${CURL_ARGS[@]}" "${BASE_PROM_URL}/api/v1.0/health"; then
+            log "Promenade API healthy"
+            break
+        elif [[ $attempt == "${MAX_HEALTH_ATTEMPTS}" ]]; then
+            log "Promenade health check failed, max retries (${MAX_HEALTH_ATTEMPTS}) exceeded."
+            exit 1
+        fi
+        sleep 10
+    done
 
-    log "Fetching join script"
+    JOIN_CURL_URL="$(render_curl_url "${NAME}" "${LABELS[@]}")"
+    log "Fetching join script via: ${JOIN_CURL_URL}"
     ssh_cmd "${VIA}" curl "${CURL_ARGS[@]}" \
-        "$(render_curl_url "${NAME}" "${LABELS[@]}")" > "${SCRIPT_DIR}/join-${NAME}.sh"
+        "${JOIN_CURL_URL}" > "${SCRIPT_DIR}/join-${NAME}.sh"
 
     chmod 755 "${SCRIPT_DIR}/join-${NAME}.sh"
     log "Join script received"

tools/gate/config-templates/site-config.yaml

@@ -39,6 +39,8 @@ data:
     insecure-registries:
       - 10.24.20.19:30092
     live-restore: true
+    max-concurrent-downloads: 10
+    oom-score-adjust: -999
     storage-driver: overlay2
 ---
 schema: promenade/HostSystem/v1
@@ -114,6 +116,7 @@ data:
     - --eviction-max-pod-grace-period=-1
     - --network-plugin=cni
     - --node-status-update-frequency=5s
+    - --serialize-image-pulls=false
     - --v=5
   images:
     pause: gcr.io/google_containers/pause-amd64:3.0