Send telemetry from your code

Monitor already reads every canister’s health, cycles and memory. To see what your app is doing, such as queue depth, failed payments or slow requests, send your own telemetry. It lands next to Monitor’s data, in the same charts, dashboards and alerts.

Pick what to send

You want	Send	Guide
Numbers over time: counts, sizes, durations	Metrics	Motoko · Rust
Searchable, structured lines	Logs	Logs from your code
One request followed across canisters	Traces	Traces from your code
Metrics from an app that already uses OpenTelemetry	Metrics	OpenTelemetry metrics

How it works

1. Add your canister as a sender of the right data connection: Telemetry › Data connections, then Senders › Add a sender. You need to be an Admin.

2. Add the client library to your canister. The sender’s guide in icLens shows the exact setup for that canister.

3. Check it arrives. The sender shows Receiving, and your metric appears in Telemetry › Metrics.

The client libraries ship with icLens’s source. They aren’t on Mops or crates.io yet, so vendor them into your project.

Choose good metrics

Start with three or four that answer a real question:

Question	Metric
Is work piling up?	A gauge of queue depth
Is work finishing?	A counter of completed jobs, split by outcome
How long does it take?	A histogram of durations

Keep labels to small, fixed sets, such as operation or outcome. Never put user IDs, request IDs or secrets in labels.

Good to know

• Your metrics can’t pose as Monitor’s: the Agent marks where every value came from.
• If a send times out, the client keeps the batch and checks before sending again. Data is never sent twice or silently dropped.
• The exact methods and types are in the SDK APIs reference.

Next

• Metrics from Motoko
• Metrics from Rust