A deployed app can fail in a few distinct ways: the platform refuses the request (a paused or gated app), the edge cannot reach what renders the page (an origin or Worker), or your own code errors. The response body usually tells you which. Match it below.
Where to look
- Traffic & analytics shows requests by status code, so you can see when errors started and which paths return them. See Traffic & analytics.
- Build & deploy logs, then What your app is printing, shows what a container app printed to stdout and stderr: stack traces, boot errors, framework logs.
- Deploys shows whether the error started with a particular deployment. Roll back from there if it did. See Deployments.
"This site is paused" (503)
The organization has no active plan: the subscription lapsed or its first payment failed. Every site in the organization serves this page, container apps sleep, and queue workers stop.
How you resume depends on why the organization was paused; Paused accounts has the steps for each case. Sites come back within the hour after payment, usually right away.
"This app is paused. The workspace reached its usage limit." (503)
The organization reached its spending cap: Starter's ($25 by default) or one you set on Pro or Team. dply stops starting the app's container. Within the hour the whole organization is paused and its sites show "This site is paused" instead. It resumes at the next billing period, or as soon as you raise the cap. See Spending caps & alerts.
"We'll be right back." (503)
Maintenance mode is on for the site. Turn off Maintenance mode under Error pages, or remove maintenance from dply.yaml and redeploy. See Error pages.
"Service temporarily unavailable — The site origin did not respond." (503)
A hybrid site proxied a request to its origin and the origin failed or timed out, twice. The response carries X-Dply-Edge-Failover: 1.
- Check the origin is up and answers at the Origin URL on Delivery. Test origin probes it.
- If the origin sits behind an access gateway or checks a shared secret, confirm Origin access token or Origin auth secret match.
- Replace the default page with your own under Failover HTML.
"Service temporarily unavailable — SSR worker not reachable." (503)
The edge could not hand the request to the site's Worker SSR script. Redeploy; if it persists, contact Support with the site's link, since this is a platform-side fault.
500 or 502 from a container app
Your app answered with an error, or crashed. Open the container logs.
| What you see in the logs | Fix |
|---|---|
| A stack trace for a missing environment variable or config | Add it under Environment and redeploy. |
SQLSTATE connection errors |
Check a database is attached under Add resource, Database, and that you redeployed after attaching it. A sleeping database wakes on the next connection; the first query waits for it. |
| Missing table errors | Migrations have not run. Turn on Run migrations when a container starts in Overview → App card → Sleep, scaling, region, or run Migrate from the database's settings. |
| Out of memory, or the container restarting | Choose a larger size on the App card on Overview. |
| Laravel: "No application encryption key" | Restore APP_KEY under Environment; dply generates one only when none is set. |
After each container deploy dply requests the app's URL and marks the deploy failed if it answers with a 5xx. See Troubleshooting builds.
The first request is slow
Container apps sleep after the Sleep after idle period (5 minutes by default). The next request starts a container and waits for it to boot, which is the cold start. It is longer when Run migrations when a container starts is on, because every start runs migrations first.
- Set Always awake to 1 in Overview → App card → Sleep, scaling, region to keep one instance warm. It bills compute all month.
- Use a longer Sleep after idle for fewer cold starts.
- Use Scaling windows to keep instances warm only during business hours.
See Scaling & sleep.
Assets load over HTTP, or styles are missing
- Laravel: mixed content. Asset links built for
http://are blocked in the browser. dply setsASSET_URLto the app's address on the first deploy; after adding a custom domain, updateAPP_URLandASSET_URLunder Environment and redeploy. - Rails: no styles.
assets:precompilefailed during the build without stopping it. Look for its output in the build log. - Static sites: 404 on client-side routes. Turn on SPA fallback under Build, Advanced.
Users are logged out, or data disappears
- Sessions. Laravel apps default to
SESSION_DRIVER=cookie, which works across instances. File sessions do not: each container has its own disk. - Uploaded files. Files written inside the container are lost when it sleeps or is replaced. Store them in Object storage.
- SQLite. An app without a database runs on SQLite inside the container. Attach a database for data you need to keep.
WebSockets
| Symptom | Fix |
|---|---|
| Echo never connects | Redeploy after adding Realtime: the VITE_* keys are baked in at build time. Check Allowed origins on the Realtime resource includes your site. |
| Broadcasts are refused after a secret rotation | Redeploy so the app signs with the new secret. |
| Connections refused at a steady number | You reached Max connections. Raise it within your plan's cap. |
| "WebSocket upgrade failed at origin." (502) | A hybrid origin did not accept the WebSocket upgrade. Use a Realtime resource, or check the origin's WebSocket support. |
See Laravel broadcasting with Realtime.