# Monitoring

Monitoring shows how your deployed apps are behaving - CPU and memory usage, request metrics, and live container logs - so you can spot problems early and understand load. It complements the per-deployment logs you get during a release.

## What you can see

- **Resource metrics** - CPU and memory per container over time, so you can see saturation and trends.
- **Request metrics** - traffic and response patterns over a selectable window (e.g. 5m, 1h, 24h).
- **Container logs** - live stdout/stderr from the running app.

## Where logs live

orkestr keeps logs close to the thing that produced them:

- **Runtime logs** - on the Monitoring page and the project, for the running app.
- **Build and deploy logs** - on each [deployment](https://orkestr.eu/docs/features/deployments), including per-step output.
- **Command output** - on each [job](https://orkestr.eu/docs/features/jobs) run, with the exit code.
- **Function logs and metrics** - on each [function](https://orkestr.eu/docs/features/functions).

## Notifications and alerts

orkestr surfaces notifications for events like failed deploys and high CPU or memory pressure, shown in the console's notification center. The Pro plan adds a monitoring dashboard, and Enterprise adds alerting so you hear about problems without watching the dashboard.

## Reading the signals

- **Memory climbing to the cap then restarts** - the app likely needs a larger per-environment memory allocation (available on paid plans), or it has a leak.
- **CPU pinned at the limit** - the workload needs more CPU headroom, or there's a hot loop.
- **5xx right after a deploy** - check the deployment's logs; it's usually a missing env var or a port mismatch.
- **First request slow after idle** - on free/Pro that's a cold start after auto-sleep, not an error.

> **Tip**
>
> High memory immediately after a deploy often just means the app needs a larger memory allocation. Raise the environment's memory toward your plan's ceiling rather than chasing a phantom leak.

## FAQ

### How far back do metrics go?

Pick a time window on the page (for example 5 minutes up to 24 hours) to zoom in or out.

### Can I get alerted automatically?

Notifications appear in the console for key events; Enterprise adds alerting on top.

### Where do I debug a crash loop?

Start with the container logs here, then the most recent [deployment](https://orkestr.eu/docs/features/deployments) to see whether it failed at build or at runtime.