From 7532e20d1e7c961ea82d6fef96781e90ae54524a Mon Sep 17 00:00:00 2001
From: Jonathan <claude.cs@jger.nl>
Date: Fri, 10 Jul 2026 17:32:20 +0200
Subject: [PATCH] fix: worker waits for schema before claiming; npm ci for
 reproducible builds

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---
 web/Dockerfile             |  2 +-
 worker/lyra_worker/db.py   | 32 ++++++++++++++++++++++++++++++++
 worker/lyra_worker/main.py |  4 ++--
 worker/tests/test_db.py    | 11 +++++++++++
 4 files changed, 46 insertions(+), 3 deletions(-)
 create mode 100644 worker/tests/test_db.py

diff --git a/web/Dockerfile b/web/Dockerfile
index 7e4782a..132e72a 100644
--- a/web/Dockerfile
+++ b/web/Dockerfile
@@ -4,7 +4,7 @@ RUN apt-get update && apt-get install -y openssl && rm -rf /var/lib/apt/lists/*
 COPY package.json package-lock.json* ./
 # Copy schema before install so package.json's postinstall (prisma generate) can run
 COPY prisma ./prisma
-RUN npm install
+RUN npm ci
 COPY . .
 RUN npx prisma generate && npm run build
 
diff --git a/worker/lyra_worker/db.py b/worker/lyra_worker/db.py
index 8c89282..e3e882f 100644
--- a/worker/lyra_worker/db.py
+++ b/worker/lyra_worker/db.py
@@ -1,4 +1,5 @@
 import os
+import time
 
 import psycopg
 
@@ -6,3 +7,34 @@ import psycopg
 def connect() -> psycopg.Connection:
     dsn = os.environ["DATABASE_URL"]
     return psycopg.connect(dsn)
+
+
+def wait_for_db(timeout: float = 60.0, interval: float = 2.0) -> psycopg.Connection:
+    """Block until the database is up and the `Job` table exists.
+
+    Repeatedly opens a connection and probes for the `Job` table's presence.
+    Tolerates the DB not being reachable yet (`psycopg.OperationalError`) and
+    migrations not having run yet (`psycopg.errors.UndefinedTable`), retrying
+    with a fixed `interval` until `timeout` elapses. Returns the open, usable
+    connection on success; re-raises the last error on timeout.
+    """
+    deadline = time.monotonic() + timeout
+    printed_waiting = False
+    last_error: Exception | None = None
+    while time.monotonic() < deadline:
+        conn: psycopg.Connection | None = None
+        try:
+            conn = connect()
+            with conn.cursor() as cur:
+                cur.execute('SELECT 1 FROM "Job" LIMIT 0')
+            return conn
+        except (psycopg.OperationalError, psycopg.errors.UndefinedTable) as exc:
+            last_error = exc
+            if conn is not None:
+                conn.close()
+            if not printed_waiting:
+                print("worker: waiting for database schema...", flush=True)
+                printed_waiting = True
+            time.sleep(interval)
+    assert last_error is not None
+    raise last_error
diff --git a/worker/lyra_worker/main.py b/worker/lyra_worker/main.py
index bbfe538..19f897b 100644
--- a/worker/lyra_worker/main.py
+++ b/worker/lyra_worker/main.py
@@ -1,7 +1,7 @@
 import time
 
 from lyra_worker.claim import claim_next
-from lyra_worker.db import connect
+from lyra_worker.db import wait_for_db
 from lyra_worker.pipeline import run_pipeline
 
 IDLE_SLEEP = 2.0
@@ -9,7 +9,7 @@ STAGE_DELAY = 1.0  # visible progress in the UI
 
 
 def run_forever() -> None:
-    conn = connect()
+    conn = wait_for_db()
     print("worker: waiting for jobs", flush=True)
     try:
         while True:
diff --git a/worker/tests/test_db.py b/worker/tests/test_db.py
new file mode 100644
index 0000000..7c2f788
--- /dev/null
+++ b/worker/tests/test_db.py
@@ -0,0 +1,11 @@
+from lyra_worker.db import wait_for_db
+
+
+def test_wait_for_db_returns_usable_connection_when_schema_ready(conn):
+    # `conn` fixture guarantees the DB is up and the schema is migrated.
+    result = wait_for_db(timeout=5)
+    try:
+        with result.cursor() as cur:
+            cur.execute('SELECT 1 FROM "Job" LIMIT 0')
+    finally:
+        result.close()