Quanamo — deploy

quanamo-deploy ships the latest tagged code for the Quanamo feedback platform to the odin production server. Three Laravel apps share the box — admin.quanamo.com, feedz.quanamo.com, kanban.quanamo.com — each with its own deploy.sh.

Unlike Kraite, Quanamo has exactly one server. There's no fleet orchestration: iterate the three apps in a fixed dependency order, gate on health, recurse into warmup.

Server

FieldValue
Host204.168.242.59
Userodin (SSH directly as odin, NOT root)
SSH key~/.ssh/id_ed25519_waygou
admin/home/odin/admin.quanamo.com — PostgreSQL master, runs core migrations
feedz/home/odin/feedz.quanamo.com — shares admin's PostgreSQL DB, no own migrations
kanban/home/odin/kanban.quanamo.com — owns its own MySQL DB (kanban_quanamo)

All commands run as odin. No su needed — odin IS the project user. System commands use sudo.

Invocation

UsageBehaviour
/do quanamo-deployFull sequence: admin → feedz → kanban → health → warmup
/do quanamo-deploy adminadmin only
/do quanamo-deploy feedzfeedz only
/do quanamo-deploy kanbankanban only

Sequence

StepAction
1. Determine tag`git tag --sort=-v:refname
2. Deploy adminDEPLOY_TAG=v<version> bash deploy.sh — runs core migrations against PostgreSQL
3. Deploy feedzSame shape — no migrations, shares admin's DB
4. Deploy kanbanSame shape — its own MySQL migrations, its own db-backups/
5. Health checkartisan version, DB connection per app, package versions, no dev-*
6. WarmupRecurse into quanamo-warmup

DB backup hard gates (per app that owns a DB)

AppBackup commandPath
admin (PostgreSQL)`pg_dump -U odin quanamogzip`
kanban (MySQL)`mysqldump -u odin kanban_quanamogzip`

The dump runs BEFORE php artisan migrate --force. If the dump exits non-zero OR produces a snapshot smaller than 1KB, the deploy aborts and migrations are NEVER executed. Old backups are never deleted — full point-in-time history sits in db-backups/ for rollback.

Hard rules

  • Core migrations run ONLY on admin. feedz shares the same PostgreSQL DB and never runs its own migrations.
  • kanban migrations run on kanban. Own DB, own migration history.
  • APP_ENV=production on odin.
  • Tag BEFORE deploy. Always run quanamo-tag first.
  • deploy.sh runs as odin — no su needed.
  • No dev-master in production.
  • A failure on kanban must NOT roll back admin or feedz, and vice-versa. Each app has its own deploy.sh and its own backup.
  • kanban active.flag is package-managed (step-dispatcher v1.12+). Don't remove it during deploy. Don't re-arm it during warmup. php artisan down is enough to pause inbound enqueues.

Rollback from backup

Get Bruno's explicit approval before restoring — a restore wipes everything that landed AFTER the backup was taken (any orders / positions / feedback the system recorded since).

AppRestore
admin (PostgreSQL)`gunzip -c db-backups/pre-deploy-<TIMESTAMP>.sql.gz
kanban (MySQL)`gunzip -c db-backups/pre-deploy-<TIMESTAMP>.sql.gz