From 5a44afe86ec30074581c65049ee17edd8b2981c2 Mon Sep 17 00:00:00 2001
From: Bryan Van Deusen <bryan.vandeusen@fabledsword.com>
Date: Thu, 8 Oct 2026 20:36:25 -0400
Subject: [PATCH] chore(compose): no healthcheck on the database (#5275)

#5275 "steward_db killed by its healthcheck twice": on NFS, pg_isready
timed out during slow writes and the orchestrator killed a healthy
database. The operator removed it from the live stack and ruled that
Steward doesn't ship one. Steward waits for the database itself (60 s at
startup, retries per request), so depends_on needs no health condition.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 compose.deploy.yml | 11 ++++-------
 docker-compose.yml | 11 ++++-------
 2 files changed, 8 insertions(+), 14 deletions(-)

diff --git a/compose.deploy.yml b/compose.deploy.yml
index c5bd8fb..3aa9050 100644
--- a/compose.deploy.yml
+++ b/compose.deploy.yml
@@ -39,8 +39,7 @@ services:
       # mount, pin STEWARD_SECRET_KEY (env or a docker secret) instead, so the key
       # never depends on filesystem ownership being correct on every node.
     depends_on:
-      db:
-        condition: service_healthy
+      - db
     restart: unless-stopped
 
   db:
@@ -54,11 +53,9 @@ services:
     volumes:
       - pgdata:/var/lib/postgresql/data
     # No host port published — db is reachable only on the backend network.
-    healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U steward"]
-      interval: 5s
-      timeout: 5s
-      retries: 5
+    # No healthcheck: a slow disk (NFS) failed pg_isready and the orchestrator
+    # killed a healthy database mid-write (#5275). Steward waits for the
+    # database itself, at startup and per request (steward/database.py).
     restart: unless-stopped
 
 networks:
diff --git a/docker-compose.yml b/docker-compose.yml
index 9f7cf81..4a3c6ea 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -27,8 +27,7 @@ services:
       - STEWARD_SECRET_KEY=dev-secret-key-change-me
       - PYTHONPATH=/app
     depends_on:
-      db:
-        condition: service_healthy
+      - db
     restart: unless-stopped
 
   db:
@@ -42,11 +41,9 @@ services:
       - "5432:5432"
     volumes:
       - pgdata:/var/lib/postgresql/data
-    healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U steward"]
-      interval: 5s
-      timeout: 5s
-      retries: 5
+    # No healthcheck: a slow disk (NFS) failed pg_isready and the orchestrator
+    # killed a healthy database mid-write (#5275). Steward waits for the
+    # database itself, at startup and per request (steward/database.py).
     restart: unless-stopped
 
 volumes: