Telemetry volume
Flat pricing removes the bill surprise, not the volume surprise. A team that turns on debug logging across a fleet still wants to know before their retention window quietly starts truncating, and "how much are we sending" is the first question anyone asks when a page loads slowly.
Usage in the sidebar, or GET /api/usage/telemetry.
This is not a meter
RootTrace bills by host and by APM service, both of them counts fixed in your licence. Volume is not one of the numbers you are charged on, so nothing on this page costs you anything and nothing here throttles. It exists for capacity planning and for finding the service that started shouting, which is exactly why it can be shown at all, and why the page says so in as many words.
A service counts toward your plan while it is reporting and stops counting seven days after it goes quiet. The same service name in two environments counts once. Current usage against both caps is on the workspace settings page, not here.
What is counted
Per-signal record counts over the window, from each collection's own time field, grouped the way an operator thinks about their telemetry rather than the way it happens to be stored.
| Signal | What drives it |
|---|---|
| Collector checks | Check cadence and how many checks are enabled per host |
| APM | Transaction throughput, custom metric cardinality, trace sampling |
| Application logs | Log level and volume from the SDKs, file tailer, and OTLP |
| Availability history | How often checks change state: a flapping check writes far more |
| Issues and incidents | Alert volume. A high count here usually means noisy thresholds |
| Audit and Linux audit | auditd rule breadth on the hosts that ship it |
The driver column is the point. "You are sending a lot of logs" is not actionable; "log level and volume from the SDKs" tells you what to change.
Byte figures are estimates
Volume is estimated from each collection's average object size, not measured. MongoDB knows the average cheaply; the exact per-document total only by reading every document, which is precisely the expensive scan this page helps people avoid. The response labels them as estimates and so does the UI.
Retention projection
Each signal reports where its current rate lands once retention fills: records-per-day × retention days. Unlimited retention has no steady state, so it reports the daily rate instead of a fabricated total.
Top contributors
GET /api/usage/top-contributors?signal=logs names the services or hosts producing the volume, with each one's share. Every signal has a different natural "who is responsible" dimension: service for logs and APM, host for checks and audit, check type for issues.
API
| Endpoint | Purpose |
|---|---|
GET /api/usage/telemetry | Volume by signal, with retention projection |
GET /api/usage/top-contributors | Which service or host is producing it |
Both take window of 24h, 7d, or 30d.