Metrics
Charts of memory, CPU, bandwidth and response time for the last 24 hours.
Web services, databases and key value stores have a Metrics page. Use it to see how much of its memory and CPU a service uses, and whether a deploy changed that. Every chart covers the last 24 hours, from samples taken every 3 minutes.
Each chart shows its Peak value and when it happened.
Web services
| Chart | What it shows |
|---|---|
| Memory Usage | Memory the service uses, as a share of the memory you gave it. |
| CPU Usage | CPU the service uses, as a share of all the machine's cores, which is the most it can use. |
| Bandwidth | Traffic in and out of the service, in KB/s. |
| Response Time (p95) | The time within which 95% of requests were answered. |
Deploys appear on the charts as markers with the commit, so you can see whether a change affected memory, CPU or response times.
Bandwidth and response time count requests that come in through the service's address and custom domains. Requests from other services in the project over the private network aren't counted.
Databases
The CPU Usage and Memory Usage charts show what the database uses. Active connections and disk usage aren't collected yet; the database's Overview shows its storage use. See Postgres storage.
Key value stores
The CPU Usage and Memory Usage charts show what the store uses. Connected clients and operations per second aren't collected yet.
Workers
Workers don't have a Metrics page. Their card in the project diagram, on the project's Services page, shows current CPU and memory use.
New services
A new service shows No data yet until its first sample arrives, usually within about a minute. Bandwidth and response time need two samples, so they start a few minutes later.
Metrics and billing
What a service uses doesn't change what you pay. Billing depends only on your machines and the services you have. See How billing works.
To see how a machine's memory, CPU and disk are shared between services, open the project's Resources page.