Refine probes based on k3d cluster verification

Every chart was installed twice in a throwaway k3d cluster (origin/main
vs this branch) with real Postgres/Redis/dex instances and observed for
readiness ordering, stability and dependency-outage behavior. Findings:

- voucher-vault: Django only accepts Hosts 127.0.0.1 and the configured
  DOMAIN; the default probe Host (pod IP) always got 400. Probes now
  send 'Host: 127.0.0.1'.
- zipline: /api/healthcheck includes the database, so liveness now uses
  '/' (readiness/startup keep /api/healthcheck). Verified live: on a
  Postgres outage the pod turns NotReady within 30s and is never
  restarted by the kubelet.
- outline: '/' also returns 500 when the database is down, so there is
  no dependency-free endpoint; all three probes use /_health.

.watch-pr.sh

@@ -0,0 +1,15 @@
+#!/bin/bash
+# Throwaway PR watcher (not committed). Exits when the PR leaves OPEN state
+# or any review/issue comment appears.
+REPO=jk-cluster/helm-charts
+PR=21
+while true; do
+  STATE=$(gh pr view $PR --repo $REPO --json state --jq .state 2>/dev/null)
+  RC=$(gh api repos/$REPO/pulls/$PR/comments --jq length 2>/dev/null)
+  IC=$(gh api repos/$REPO/issues/$PR/comments --jq length 2>/dev/null)
+  if [ -n "$STATE" ] && [ "$STATE" != "OPEN" ]; then break; fi
+  if [ -n "$RC" ] && [ "$RC" != "0" ]; then break; fi
+  if [ -n "$IC" ] && [ "$IC" != "0" ]; then break; fi
+  sleep 90
+done
+echo "STATE=$STATE REVIEW_COMMENTS=$RC ISSUE_COMMENTS=$IC"

outline/templates/outline.sfs.yaml

@@ -80,8 +80,10 @@ spec:
             - containerPort: 3000
               protocol: TCP
               name: http
-          # Outline exposes a dedicated health endpoint at /_health
-          # (also used by the upstream Docker HEALTHCHECK).
+          # Outline exposes a dedicated health endpoint at /_health (also used
+          # by the upstream Docker HEALTHCHECK). It checks Postgres and Redis;
+          # there is no dependency-free HTTP endpoint ("/" returns 500 too when
+          # the database is down), so all three probes use /_health.
           livenessProbe:
             httpGet:
               path: /_health

voucher-vault/values.yaml

@@ -13,11 +13,17 @@ voucherVault:
       cpu: 1
       memory: 384Mi
       ephemeral-storage: 100Mi
-  # "/" answers with a redirect for anonymous users, which httpGet treats as success.
+  # "/" answers with a redirect for anonymous users, which httpGet treats as
+  # success. The Host header must be set to 127.0.0.1 because Django's
+  # ALLOWED_HOSTS only accepts 127.0.0.1 and the configured DOMAIN — probing
+  # with the default Host (the pod IP) would always return 400.
   livenessProbe:
     httpGet:
       path: /
       port: http
+      httpHeaders:
+        - name: Host
+          value: 127.0.0.1
     periodSeconds: 10
     timeoutSeconds: 5
     failureThreshold: 3
@@ -25,6 +31,9 @@ voucherVault:
     httpGet:
       path: /
       port: http
+      httpHeaders:
+        - name: Host
+          value: 127.0.0.1
     periodSeconds: 10
     timeoutSeconds: 5
     failureThreshold: 3
@@ -33,6 +42,9 @@ voucherVault:
     httpGet:
       path: /
       port: http
+      httpHeaders:
+        - name: Host
+          value: 127.0.0.1
     periodSeconds: 5
     timeoutSeconds: 5
     failureThreshold: 30

zipline/values.yaml

@@ -8,10 +8,12 @@ config:
   maxUploadFileSize: 30m
 zipline:
   # Zipline v4 exposes /api/healthcheck (checks server and database health);
-  # the endpoint is enabled by default (config.features.healthcheck).
+  # the endpoint is enabled by default (config.features.healthcheck). Since it
+  # includes the database, it is used for readiness/startup only; liveness
+  # uses "/" so a database outage does not restart the pod.
   livenessProbe:
     httpGet:
-      path: /api/healthcheck
+      path: /
       port: zipline-server
     periodSeconds: 10
     timeoutSeconds: 5