Skip to content
Weik.ioDocs
2026.2
Try Weik.io

Watching the calls

Weik.io 2026.2

The gateway logs each request with its trace, splits failures into the server's side and the caller's, and every response's traceresponse header finds the call. Product page · Video

Watch: Watch the calls, 3:14

The video finds each call on the Requests page and follows one of them into its trace.

On this page

Every gateway instance counts every call and records whose side a failure is on. The server’s side is when the gateway or the system behind it failed. The caller’s side is a missing or unknown credential, a subscription that doesn’t open the API or has expired, a rate limit or a malformed request. The figures, the requests and the traces are on the API’s own page and in the CLI.

The Requests page#

An API’s page has Requests under Observability, with one row per call: the time, the method and path, the version, the subscription, the status and what happened. You filter by version, subscription, route, status, whose side failed or slower than a time you set, and a row opens the call’s trace through the gateway into the integration behind it.

/apis/finance/invoices/requests
The Requests page of the invoices API in production with filters, a chart of calls and one row for each call from the subscriptions ap-portal, partner-x and partner-yThe Requests page of the invoices API in production with filters, a chart of calls and one row for each call from the subscriptions ap-portal, partner-x and partner-yThe Requests page of the invoices API in production with filters, a chart of calls and one row for each call from the subscriptions ap-portal, partner-x and partner-yThe Requests page of the invoices API in production with filters, a chart of calls and one row for each call from the subscriptions ap-portal, partner-x and partner-y
Requests has one row for each call through the gateway, with the subscription, the path, the status and the time it took, and each row opens the trace of the call.

The figures#

The API’s Overview has a strip for the last day: calls and requests a second now, the share that failed on the server’s side and on the callers’ side, the response time and the part of it spent in the integration behind. Metrics shows four numbers and two charts, with tabs for status codes, versions, routes and subscriptions. The versions tab shows the calls, failures and last call of each version, so you see when an old version can be retired. The routes tab says when a declared route had no calls.

From the CLI:

Terminal
$ weikio metrics --api finance/invoices --env production --by status
$ weikio metrics --api finance/invoices@v1 --env production --by subscription

Find one call#

Every response the gateway sends carries a traceresponse header. Ask a caller who reports a failed call for that value:

Terminal
$ curl -i -H "X-Api-Key: $PARTNER_KEY" https://api.example.com/invoices/v1/pending

Paste the value into the search of Observability › Traces, or give it to the CLI, and you get the call’s whole path: the gateway, the route it matched and every step of the integration behind it.

Terminal
$ weikio trace 00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01

A trace started at the gateway shows as the API it called. A trace that crosses into a project you can’t read shows the steps you may see and says how many it hides.

Logs of the gateway#

The API’s Logs section lists the lines the gateway instances write for the API in one environment. Observability › Logs has an API filter beside its other filters, and its Runs on column says APIM instance for a line the gateway wrote.

When an API fails#

An API that fails on the server’s side for 5 % of its calls in 15 minutes raises a problem. The callers’ own mistakes don’t count. The problem shows in Home’s Needs attention, on the Observability Overview, on the API’s page and in weikio problems, with a few words of figures such as 42 of 310 failed in 15 min.

The gateway keeps every trace of a call until it is 7 days old or the trace store reaches its disk cap. A busy installation can trace only a share of the calls, with WEIKIO_APIM_TRACE_SAMPLE_RATIO in the hub’s .env, and a call’s trace is then kept whole or not at all.