From 9dd2a720636461b267b03ed4a8034b283ba8fc21 Mon Sep 17 00:00:00 2001
From: Brad Stein <Brad.Stein@gmail.com>
Date: Sat, 17 Jan 2026 02:16:13 -0300
Subject: [PATCH] mailu: retry sync and rerun job

---
 services/mailu/mailu-sync-job.yaml   |  2 +-
 services/mailu/scripts/mailu_sync.py | 28 +++++++++++++++++++++++++---
 2 files changed, 26 insertions(+), 4 deletions(-)

diff --git a/services/mailu/mailu-sync-job.yaml b/services/mailu/mailu-sync-job.yaml
index 6da615e..18aef7c 100644
--- a/services/mailu/mailu-sync-job.yaml
+++ b/services/mailu/mailu-sync-job.yaml
@@ -2,7 +2,7 @@
 apiVersion: batch/v1
 kind: Job
 metadata:
-  name: mailu-sync-6
+  name: mailu-sync-7
   namespace: mailu-mailserver
 spec:
   template:
diff --git a/services/mailu/scripts/mailu_sync.py b/services/mailu/scripts/mailu_sync.py
index 74b170a..d1754cb 100644
--- a/services/mailu/scripts/mailu_sync.py
+++ b/services/mailu/scripts/mailu_sync.py
@@ -42,6 +42,28 @@ def log(msg):
     sys.stdout.flush()
 
 
+def retry_request(label, func, attempts=10):
+    for attempt in range(1, attempts + 1):
+        try:
+            return func()
+        except requests.RequestException as exc:
+            if attempt == attempts:
+                raise
+            log(f"{label} failed (attempt {attempt}/{attempts}): {exc}")
+            time.sleep(attempt * 2)
+
+
+def retry_db_connect(attempts=10):
+    for attempt in range(1, attempts + 1):
+        try:
+            return psycopg2.connect(**DB_CONFIG)
+        except psycopg2.Error as exc:
+            if attempt == attempts:
+                raise
+            log(f"Database connection failed (attempt {attempt}/{attempts}): {exc}")
+            time.sleep(attempt * 2)
+
+
 def get_kc_token():
     resp = SESSION.post(
         f"{KC_BASE}/realms/{KC_REALM}/protocol/openid-connect/token",
@@ -175,13 +197,13 @@ def ensure_mailu_user(cursor, email, password, display_name):
 
 
 def main():
-    token = get_kc_token()
-    users = kc_get_users(token)
+    token = retry_request("Keycloak token", get_kc_token)
+    users = retry_request("Keycloak user list", lambda: kc_get_users(token))
     if not users:
         log("No users found; exiting.")
         return
 
-    conn = psycopg2.connect(**DB_CONFIG)
+    conn = retry_db_connect()
     conn.autocommit = True
     cursor = conn.cursor(cursor_factory=RealDictCursor)