Add-deployment-scaling-notes
This commit is contained in:
+3
-3
@@ -9,7 +9,8 @@
|
||||
"php": "^8.3",
|
||||
"laravel/framework": "^13.8",
|
||||
"laravel/sanctum": "^4.3",
|
||||
"laravel/tinker": "^3.0"
|
||||
"laravel/tinker": "^3.0",
|
||||
"spatie/laravel-permission": "^8.3"
|
||||
},
|
||||
"require-dev": {
|
||||
"fakerphp/faker": "^1.23",
|
||||
@@ -19,8 +20,7 @@
|
||||
"laravel/pint": "^1.27",
|
||||
"mockery/mockery": "^1.6",
|
||||
"nunomaduro/collision": "^8.6",
|
||||
"phpunit/phpunit": "^12.5.12",
|
||||
"spatie/laravel-permission": "^8.3"
|
||||
"phpunit/phpunit": "^12.5.12"
|
||||
},
|
||||
"autoload": {
|
||||
"psr-4": {
|
||||
|
||||
@@ -43,6 +43,18 @@ Provider Exchange is a Laravel application for a provider-centered field-service
|
||||
9. Admin operations: audit logs, queue health, system status, content moderation, and backup/export workflows.
|
||||
10. Deployment hardening: production environment docs, queue worker setup, scheduler setup, storage policy, and server provisioning notes.
|
||||
|
||||
## Scaling Risks To Track
|
||||
|
||||
- Search and filtering will need proper database indexes first, then a dedicated search service if directory/job volume grows.
|
||||
- Ratings and reputation pages should eventually use cached aggregates instead of recalculating averages and counts on every page view.
|
||||
- Notifications should move from the database queue to Redis-backed queues before high-volume alerting or email delivery.
|
||||
- Sessions, cache, and queues should move to Redis before running more than one web node.
|
||||
- Uploaded evidence, profile images, and deliverable files should move to shared or object storage before multi-server deployment.
|
||||
- Work-order timelines, dispute activity, messages, and notification tables need careful indexes before large production use.
|
||||
- Realtime chat, websocket presence, mobile push, and geolocation check-ins should be added as deliberate service layers instead of bolted onto normal page requests.
|
||||
- Background imports, image processing, moderation summaries, and reputation recalculation should run through queue workers rather than web requests.
|
||||
- Production deployments need explicit PHP-FPM pool, Nginx, scheduler, queue worker, backup, and log rotation documentation.
|
||||
|
||||
## Future/Deferred
|
||||
|
||||
- Escrow or payment custody.
|
||||
|
||||
Reference in New Issue
Block a user