Traffic and consumption

The project's Observability page brings together what the platform measures about your application, in five tabs.

Tab	What it answers
Overview	The period at a glance: volume, errors, latency, and consumption side by side
Requests	Requests over time, split by response class
Runtime	CPU, memory, and ready instances
Routes	Which routes receive traffic, with the error rate and latency of each
Errors	Error responses only, by route

Period

Last hour, 6 hours, 24 hours, 7 days, or 30 days. The choice applies to every tab.

When a tab says "No requests in this period", widening the period is the first thing to try: internal applications often have long stretches with no traffic.

Traffic

Requests are grouped by response class — success, client error, server error — rather than code by code. That is the distinction that matters day to day: a rise in client errors is usually a misconfigured integration; a rise in server errors is the application.

Latency appears as p95: the time below which 95% of responses landed. An average hides exactly the tail that bothers users.

Routes

The route table shows volume, error rate, and p95 per route — and it is usually where investigation starts: a 2% error rate overall can be 100% on one specific route.

Routes are normalized: parts that vary per request (identifiers) are grouped, so a route does not become one series per customer. When there are too many routes, the least used are summed into others.

Runtime

CPU and memory consumption over time, and how many instances were ready and unavailable. It is the reading that tells you whether a latency spike came from a resource shortage or from something else.

To adjust what the service reserves, see Resources and limits.

When there is no data

What the screen says	What it means	What to do
Observability is not collecting yet	The measurement path is not active for this project	The screen names the step — usually, confirming the public address
No requests in this period	It is measuring and nobody visited	Widen the period
No runtime samples in this period	Consumption collection produced no sample	See What exists today
No error responses in this period	There was traffic and no errors	Nothing

That last line is the distinction this screen exists to preserve: "no errors" and "measured nothing" never appear with the same text.

Next steps

• Logs
• Resources and limits
• Common problems