Files
openobserve/tests/ui-testing/pages/rumPages
Bhargav 47d510aa16 feat: rebuild the RUM error detail page around issue impact (#13697)
Design at: designs/mobile-rum-sdk-evaluation/design-docs
The page showed a single error event: header, tag strip, stack trace,
replay button, event list. It could not say how bad the error was, who
it reached, whether it was new, or where it concentrated — it never
queried anything beyond that one row.

It is now a two-column triage surface: the diagnosis path (occurrences
chart, stack trace, breadcrumbs) on the left, corroborating evidence
(replay, backend trace, breakdown, who hit it) on the right.

- ErrorHeader: error type as the title with handling/source badges, the
message in an alert banner, and a context line carrying event id, route
and the service/version/env chips. Copy-permalink and copy-event-id
actions.
- ErrorImpactStrip: occurrences, users, sessions, first/last seen. A
dash, not a zero, when a figure is unavailable.
- ErrorOccurrencesChart: this signature over time, marking the bucket
that holds the event the page was opened on.
- ErrorFacetBreakdown: browser / OS / release / page as
share-of-occurrences bars, calling out any value that owns 80%+ of the
traffic — usually the fix.
- ErrorContextCard: identity, device, geo and deployment for this
occurrence. Supersedes ErrorTags, which is removed.

Backed by useErrorDetail + errorDetailQueries: an impact aggregate gates
a histogram and two facet group-bys. Three searches, staged inside the
server's work-group cap and aborted when the view is left. Facets are
split across two group-bys on purpose — one query over all four columns
returns their cross product.

Also fixes the breadcrumb trail: it read an hour of the whole stream
with no session filter, so "what the user did before the crash" could be
other people's clicks. It is now scoped to the session id.
2026-08-07 12:31:19 +00:00
..