Logs & telemetry

Logs, traces and metrics are all queryable in-product. You do not need an external observability tool open to answer why a request failed, and everything you can see is scoped to your own organization.

Where to look#

There are two altitudes. Project-level pages answer questions about one application; organization-level pages answer questions across everything you run.

Who can see what#

Project-level observability follows project access: if you can see the project, you can see its logs, metrics and traces.

The organization-level pages under /orgs/<org>/platform/ are restricted to owners and admins, and each is independently feature-gated. Logs, traces and live tail can be enabled separately, so having one does not imply having the others. When a page is unavailable you are redirected with the reason attached rather than shown an empty screen.

Searching logs#

Log search is a full-text query over the log body plus a set of structured filters, returned newest first. Results are cursor-paged: a response carries a cursor you pass to the next call to get a stable next window, rather than an offset that shifts as new logs arrive.

A page returns 40 results by default and will not return more than 200 at a time. If you are pulling a large window, page through it — asking for a bigger limit will not get you one.

Log search filters#

The logs page accepts these as query-string parameters, so a filtered view is a shareable URL. It also accepts a trace identifier, which is how the trace view jumps you to the log lines belonging to one trace.

Traces#

Trace search takes the same shape as log search — a text query over span names and attributes, plus service, status and time-range filters, with the same cursor paging and the same 40/200 limits.

Opening a trace gives you its ordered spans together with the log lines correlated to it, in one view. That correlation is the point: you do not have to copy a trace identifier into a separate log search to see what a failing span actually printed.

Live tail#

Live tail streams log lines and trace rows as they arrive, filtered by query, service, severity or namespace, and optionally starting from a point in the recent past rather than from now.

Each streamed log line carries its timestamp, service, severity, body and trace identifier — so a line that looks interesting in the stream can be taken straight into a trace drilldown.

Saved searches#

A filter combination you keep re-typing can be saved, listed and deleted from the logs page. Applying a saved search navigates to the logs page with its filters already in the URL, which is also what makes it shareable with a colleague.

Deploy logs from the CLI#

Build and deploy logs for a single deployment are available directly from the CLI. The deployment ID may be a short prefix — the same one shown in the deploy list — rather than the full identifier.

platform logs <deploymentId>
platform logs <deploymentId> --tail 100
platform logs <deploymentId> --follow

--follow stops on its own when the deployment finishes, so it is safe to use as the last step of a script without a timeout around it.