Files
Mauricio Siu c7487d8c09 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>
2026-07-08 02:44:14 -06:00
..