From c7487d8c093aea2f15c4a126d140154d27a43f06 Mon Sep 17 00:00:00 2001
From: Mauricio Siu <siumauricio@icloud.com>
Date: Wed, 8 Jul 2026 02:44:14 -0600
Subject: [PATCH] fix: db healthcheck window too short and blocking depends_on
 aborted deploy

The deploy failed with 'dependency failed to start: container db-1 is
unhealthy'. Root cause: the MariaDB 11.8 healthcheck allowed only ~40s
(start_period 10s + 3x10s retries) before flagging the container
unhealthy, which is not enough for the first-boot datadir
initialization on a modest VPS. Since create-site depended on
db:service_healthy, docker compose up -d aborted the whole deploy.

Fixes, following the pattern of the existing working erpnext template:
- Extend the db healthcheck window (start_period 120s, 30 retries) so
  first-boot initialization does not mark the container unhealthy.
- Drop depends_on from the one-shot create-site and migration jobs and
  gate create-site with wait-for-it (db/redis) inside its command, so
  docker compose up -d returns without blocking on the multi-minute
  site creation and cannot fail on transient health states.

Verified on a fresh deploy: all services up, create-site installed
frappe + erpnext and enabled the scheduler, frontend serves the
ERPNext login page with HTTP 200.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 blueprints/erpnext-v16/docker-compose.yml | 27 ++++++-----------------
 1 file changed, 7 insertions(+), 20 deletions(-)

diff --git a/blueprints/erpnext-v16/docker-compose.yml b/blueprints/erpnext-v16/docker-compose.yml
index da7ea9f1..3201b4d7 100644
--- a/blueprints/erpnext-v16/docker-compose.yml
+++ b/blueprints/erpnext-v16/docker-compose.yml
@@ -172,18 +172,12 @@ services:
       replicas: ${CREATE_SITE:-0}
       restart_policy:
         condition: none
-    depends_on:
-      db:
-        condition: service_healthy
-      redis-cache:
-        condition: service_healthy
-      redis-queue:
-        condition: service_healthy
-      configurator:
-        condition: service_completed_successfully
     entrypoint: ["bash", "-c"]
     command:
       - >
+        wait-for-it -t 300 $${DB_HOST}:$${DB_PORT};
+        wait-for-it -t 120 redis-cache:6379;
+        wait-for-it -t 120 redis-queue:6379;
         export start=`date +%s`;
         until [[ -n `grep -hs ^ sites/common_site_config.json | jq -r ".db_host // empty"` ]] && \
           [[ -n `grep -hs ^ sites/common_site_config.json | jq -r ".redis_cache // empty"` ]] && \
@@ -222,17 +216,10 @@ services:
       replicas: ${MIGRATE:-0}
       restart_policy:
         condition: none
-    depends_on:
-      frontend:
-        condition: service_healthy
-      configurator:
-        condition: service_completed_successfully
-      create-site:
-        condition: service_completed_successfully
     entrypoint: ["bash", "-c"]
     command:
       - >
-        curl -f -H "Host: ${SITE_NAME}" http://frontend:8080/api/method/ping || { echo "Site busy"; exit 0; };
+        curl -f -H "Host: ${SITE_NAME}" http://frontend:8080/api/method/ping || { echo "Site not reachable yet, skipping migration"; exit 0; };
         bench --site all set-config -p maintenance_mode 1;
         bench --site all set-config -p pause_scheduler 1;
         bench --site all migrate;
@@ -250,10 +237,10 @@ services:
         condition: always
     healthcheck:
       test: ["CMD", "healthcheck.sh", "--connect", "--innodb_initialized"]
-      start_period: 10s
-      interval: 10s
+      start_period: 120s
+      interval: 5s
       timeout: 5s
-      retries: 3
+      retries: 30
     command:
       - --character-set-server=utf8mb4
       - --collation-server=utf8mb4_unicode_ci