Skip to content

Endpoints

In DeepCube™ Web, the Endpoints page lists every route your services served in the time range, one row per route and service, with its request rate, errors, p95 and p99. After reading this page you can sort and filter the list, read each column correctly, and open a route to see its latency, errors and the traces through it.

Open Endpoints under Signals in the navigation. The page follows the time range picker; see Getting around.

The list covers every route in the whole range

The list reads every server span in the range, not only the last hour, and groups them by route and service. A route that two services serve has one row for each. The head of the page gives the count, for example "8 routes in range", or "100 of 240 routes in range" while more are still to load.

The Endpoints page listing eight routes, each with its service under the route, sorted by errors

Column What it shows
Endpoint The HTTP method and the route, with the service that served it underneath.
Req/min Requests over the range, divided by the range's length in minutes, rounded to the nearest whole minute (at least 1).
Share This route's share of the requests of the routes loaded so far, with a bar scaled to the busiest loaded route.
Errors The share of requests that failed, with a warning icon at 2 % or more and an error icon at 5 % or more.
p95 The 95th percentile duration, estimated by the store.
p99 The 99th percentile duration, estimated by the store.
Trend Requests per minute over the range, as a small line. A gap in the data is a break in the line, never a dip to 0.

A request counts as failed when its span's stage is Failed or it recorded an exception. A value nothing reported shows No data, never 0.

Share and Trend have their own limits

Share is worked out over the rows loaded so far, so it changes as more rows load. Trend is read from the request metrics by route and method, not by service, so a route that two services serve shows the same trend in both rows. When the trend cannot be read, the column says why.

Sort, filter and load more routes

  • Sort by selecting a column heading: Endpoint (by name), Req/min, Errors, p95 or p99 (each highest first). The list opens sorted by Errors. The server sorts every route in the range, so changing the sort reloads the list from the top. Share and Trend do not sort.
  • Filter with the Filter routes box. It matches the method, the route and the service name, ignoring case. The filter works only on the rows loaded so far; the line above the table says how many of the loaded routes are shown.
  • Load more by scrolling to the end of the table, or by selecting Load more routes. Each page adds up to 100 routes. If a page fails to load, the rows already shown stay, and Try again loads it again.

Open a route to see its figures and traces

Select a row, or focus it and press Enter, to open the endpoint panel beside the list. How panels stack, close and go back is described in Getting around.

The endpoint panel for POST /payments/charge, with four figures, latency and error charts, and the slowest traces through the route

The panel shows, from top to bottom:

  1. Four figures: Req/min, Errors, p95 and p99. p95 and p99 are estimated from a sample of up to 1,000 requests of the route.
  2. Most failures return a status code, when the route's errors are at 5 % or more. It names the failing status code (4xx or 5xx) most of the failed requests returned, and its share, from the metrics workspace.
  3. Latency and errors: Latency, p99 and p90 and Share of requests that failed (5xx), over the whole range, from the request metrics of the route. p95 over time is not charted yet.
  4. Slowest traces and Recent errors: up to five traces each whose root span is this route, picked from the newest 100 traces in the range that mention the route. Each shows the time, the trace ID, its stage and its duration. Select one to open its trace; see Traces.

The four figures cover at most the last hour

The figures at the top read diagnostics, which cover at most the last hour of the range. When your range is longer, the panel says Clamped to the last hour: and the window it covered. The charts and the trace lists cover the whole range. So the panel's figures can differ from the route's row in the list, which covers the whole range.

Things that change what the panel shows

  • Opened from a service panel's Endpoints tab, the panel says "Figures for this route in" followed by the service name, and its figures and charts are for that service only. See Services.
  • The charts filter by method. Only GET, POST, PUT, PATCH, DELETE, HEAD and OPTIONS routes are charted. For any other method the panel says Not served yet.
  • The two error measures differ. The Errors figure counts requests whose span failed or recorded an exception. The errors chart counts 5xx responses. The two can disagree.
  • The trace lists can be capped. When more than 100 traces in the range mention the route, the panel says older traces through the route are not listed.
  • No requests in the window. When diagnostics read no requests through the route in the last hour of the range, the figures say No requests through this route in this window, and its traces are still listed below.