stripsior committed 3 months ago · 2 changed files
- Extend the mysql healthcheck window (60s start_period, 30 retries)
so slower VPS instances have time to become healthy.
- Let the migrate container retry on failure via Docker's own
restart policy instead of dying once and blocking the whole stack.
- migrate.ts: MySQL auto-commits DDL statement-by-statement, so a
migration killed mid-file leaves some statements already applied
even though the file was never marked as migrated. Run statements
individually and treat "already exists" MySQL errors as already
applied instead of crashing the whole migrate run.