The deployment finished. The pod is running. Someone still gets an error when they open the page. All three things can be true.
At that point, the next useful question is small: what happened to that request? That is the job of Requests, added in Hakopod alpha.20. It brings completed HTTP ingress records and a service’s current routing into the dashboard, so the first investigation has somewhere concrete to begin.
Start with the path someone used
Open Requests and narrow the scope to the application. Filter by status or search for the host and path. Pick a record and look at the method, response status, duration and selected backend before jumping to the application logs.
A 404 from the expected backend leads to different questions from a request that never selected your application. Did the application recognize the path? Did the domain point at the service you meant? A record cannot answer every question, but it can remove some guesses.
The view applies your existing access boundaries. Seeing a global Requests page means seeing applications you can read. Unknown ingress backends require installation administration; a supplied Host header does not grant access to someone else’s traffic.
Two clocks on the same screen
A request record describes a completed event. The routing diagram describes the current ingress and endpoints. Keep those times separate.
If you deploy a new revision after an error, the route may now point at a different pod. The old record retains the backend that handled the original request. The diagram is useful for checking where traffic can go now; it does not claim that today’s endpoints handled yesterday’s error.
Timing fields need the same care. Some overlap, and a negative HAProxy timing is shown as “Not measured.” Adding every number together produces a confident-looking answer to the wrong question.
Enough detail without collecting the whole conversation
Requests excludes query strings, headers, cookies and bodies. Client addresses are masked before logging. That still leaves useful operational facts: method, host, path, status, bytes, timing and backend.
Paths remain visible. If your application puts a reset token or a customer’s email address in a path, that value can still appear. Keeping sensitive material out of URLs is an application decision as well as a logging decision.
Retention is bounded to at most 24 hours and the newest 100,000 records across the installation. Busy traffic can shorten that window, and collection can have gaps. Use it to investigate a problem while it is fresh, not to calculate invoices or promise a complete audit trail.
Start with the Requests guide. Find the record, inspect the route, then follow the evidence into the workload.