ObserveObservability
Observability
Production's traffic at a glance. Requests, server errors and response times over time, the busiest routes, status codes and regions, worked out from your project's runtime logs.
What it shows
Open a project and choose Observability. The page charts how production is doing: the requests Orbits' edge sent to the project's production deployments, at the project's address or a build's own address. Previews are left out.
Everything on it comes from the runtime logs, so there is nothing to install. It is worked out each time you open the page.
| Panel | What it tells you |
|---|---|
| Requests | Requests per bucket, with server errors (5xx) marked, and the total against the period before |
| Response time, p75 and p99 | How long the slower requests took, per bucket. The totals add p50 |
| By route | The 20 busiest routes, each with its requests, errors, p75 and p99. Most errors first sorts them by errors |
| Status codes | Requests by class: 2xx, 3xx, 4xx, 5xx |
| Regions | Requests by region. Abuja (abj-1) is the only one with traffic |
| Alerts | The project's alert rules and their state. See alerts |
Response times are milliseconds from the edge receiving a request to the last byte of the answer. Percentiles are exact, not estimates: p75 is the time three in four requests beat, p99 the time 99 in 100 beat.
Time range and filters
| Control | Choices |
|---|---|
| Time range | 1 hour, 24 hours (the default) or 7 days |
| Route | One of the 100 busiest routes in the range |
| Status | 2xx, 3xx, 4xx or 5xx. Narrows every panel except Status codes, so you can still see the split |
| Region | A region code |
The charts are cut into buckets on Lagos time:
| Range | Buckets |
|---|---|
| 1 hour | 12 of 5 minutes |
| 24 hours | 24 of an hour |
| 7 days | 7 of a day |
The last bucket is the one still filling, so it usually looks lower than the rest. Buckets where nothing came in show zero. The totals compare the range with the same length of time before it ("on the hour before").
The project's overview page uses the same numbers over 6 hours, in 12 buckets of 30 minutes, for its health chart.
How routes are grouped
Routes are paths with the parts that vary taken out, so one page with many ids shows as one row:
- A path segment that looks like an id becomes
[id]:/products/8412is/products/[id]. - A static folder's files are one route: everything under
/_next/is/_next/*.
Query strings are never part of a route, because they are never stored.
Who can see it
Any member of the team, viewers included. Viewers can see the Alerts card but not change it.
From the API
GET /v1/teams/{team}/projects/{project}/observability returns the page's numbers: range (1h, 6h, 24h or 7d, default 24h), and optionally route (from the answer's all_routes), status and region.
$ curl -s "https://api.theorbits.app/v1/teams/acme/projects/storefront/observability?range=7d" \ -H "Authorization: Bearer $ORBITS_TOKEN"The answer has:
| Field | |
|---|---|
totals, previous | requests, errors (5xx), p50_ms, p75_ms and p99_ms for the range, and for the period before |
buckets | Oldest first: each one's start, requests, errors, p75_ms, p99_ms |
routes | The 20 busiest, with the same figures |
statuses | Requests by 2xx, 3xx, 4xx, 5xx |
regions | Requests per region |
all_routes | The 100 busiest routes, for the route filter |
bucket_seconds, as_of | The bucket size and when the numbers were worked out |
A 503 means the logs can't be read for a moment; try again shortly. The full reference is at API › Logs.