Wait for cluster DNS before starting nginx in the frontend

deploy/helm/thecombine/charts/frontend/templates/deployment-frontend.yaml

@@ -67,6 +67,13 @@ spec:
           ports:
             - containerPort: 80
             - containerPort: 443
+          readinessProbe:
+            # The entrypoint waits for cluster DNS to resolve the backend before
+            # starting nginx, so a running container is not yet a serving one.
+            tcpSocket:
+              port: 80
+            initialDelaySeconds: 5
+            periodSeconds: 5
           resources:
             requests:
               cpu:  1m

nginx/init/05-wait-for-backend-dns.sh

@@ -0,0 +1,32 @@
+#! /bin/bash
+
+###############################################################################
+# nginx resolves the hostnames in its proxy_pass directives when it loads its
+# configuration, and exits if any of them cannot be resolved:
+#
+#   [emerg] host not found in upstream "backend"
+#
+# On a cold start the frontend container can be running before the cluster DNS
+# can answer for the backend service, so wait for the name before starting.
+###############################################################################
+
+BACKEND_HOST=backend
+MAX_WAIT_SECONDS=120
+
+if getent hosts "${BACKEND_HOST}" > /dev/null 2>&1 ; then
+    exit 0
+fi
+
+echo "Waiting up to ${MAX_WAIT_SECONDS}s for '${BACKEND_HOST}' to resolve"
+until getent hosts "${BACKEND_HOST}" > /dev/null 2>&1 ; do
+    # SECONDS is wall clock since this shell started, so the cap covers the time
+    # getent spends blocking on resolver timeouts, not just the sleeps.
+    if [ "${SECONDS}" -ge "${MAX_WAIT_SECONDS}" ] ; then
+        # Start nginx anyway so that it reports the problem itself, rather than
+        # leaving a container that is running but never serving.
+        echo "'${BACKEND_HOST}' did not resolve after ${SECONDS}s"
+        exit 0
+    fi
+    sleep 2
+done
+echo "'${BACKEND_HOST}' resolved after ${SECONDS}s"

nginx/templates/default.conf.template

@@ -1,3 +1,6 @@
+# Maintainer note: any new proxy_pass host must also be added to
+# nginx/init/05-wait-for-backend-dns.sh, else nginx will exit at startup when
+# that name is not yet resolvable.
 server {
     listen 80;
     server_name ${SERVER_NAME};