Metrics and alerts
Your app's processor, memory, requests and response time for the last 30 minutes, and alerts when it crashes, restarts often or runs out of memory.
Updated 17 Oct 20262 min read
Every app's page opens with four numbers, under Right now: how much processor and memory it uses, how many requests it answers, and how fast. Below them, alerts tell you when something goes wrong, under the bell and by email.
The four numbers
| Number | What it shows |
|---|---|
| CPU | Processor use, as a share of the app's processor limit |
| Memory | Memory held, against the app's memory limit |
| Requests | Visits answered a minute |
| Typical response | The median time a visit takes: half are faster, half slower |
Each shows the last 30 minutes, minute by minute, and refreshes every 15 seconds. A minute's numbers are kept for 24 hours.
Processor counts the web process for now. Requests and response times come from the visits in your logs, so they fill once your app has visitors.
Reading them
Memory near its limit: the app may run out at a busy moment. Raise the limit, or find what holds the memory. See Limits.
CPU at its limit: the app is slowed down to stay within it. Visits take longer; raise the processor limit if it stays there.
Typical response rising: a slow database query or an outside service is often the cause. The project's Logs tab shows how long the Slowest 5% of visits take.
Example: Asha's sale
On the morning of a sale, Asha watches duka's page. Requests climb from 40 to 400 a minute. Memory goes from 210 MB to 430 MB of its 512 MB limit, and Typical response from 90 ms to 160 ms. She chooses Change limits and moves memory to 1 GB, without a restart. After the sale, use falls back, and so does what she pays: limits cost nothing by themselves.
Alerts
| Alert | When |
|---|---|
| Keeps crashing | The app or one of its workers stopped 5 times within 10 minutes, so we stopped starting it |
| Restarts often | It restarted 5 times within an hour; at most one alert an hour |
| Ran out of memory | Its processes were stopped for using all their memory, and started again; at most one alert an hour |
| Went back to the version before | A new version stopped within 5 minutes of going live |
| Deploy didn't go live | A deploy failed; the version before keeps serving |
Every alert goes under the bell in the dashboard.
Choose how you're told
Open your Profile.
Under Alerts, find An app or one of its workers keeps crashing, An app restarts often, An app runs out of memory and A deploy didn't go live, or was rolled back.
Choose Email for each one you want by email.
Text messages aren't set up yet; when they are, you can choose Text message here too.
Want more?
Send your logs to a service with its own alerts, like Better Stack or Datadog. See Logs.