Skip to main content

Issues

Errors grouped by fingerprint — one row per real problem, not per occurrence. Four surfaces feed it: failed jobs, endpoint responses, outbound adapter calls, and browser errors reported by the client script.

The list gives Issue, Source, Events, Last seen and Status, with filters for source, status and application. Always shown — issue grouping is a Core feature, not an addon.

How things group

The fingerprint is hash(source + exception type + locus):

  • The message is normalised out of identity — digits, GUIDs, hex and quoted literals collapse to placeholders — so Order 123 not found and Order 456 not found are one issue, not two. The normalised message becomes the title, which is also what makes the title PII-safe.
  • Stack-bearing sources (jobs, client) group on the top in-app stack frame, so two bugs in one handler stay two issues.
  • Source is part of identity, so a browser TypeError never merges with a server exception.

Beyond a per-source cap (2000 by default) new groups collapse into an {other} bucket rather than growing without bound — the client source is fed by a public endpoint, so this guard matters.

What counts

SourceCounted
JobsEvery caught exception — retry attempts and terminal, so a flaky handler is visible
Endpoints5xx and unhandled exceptions, plus 4xx as separate status-code groups (filtered out by default, and kept off the error-rate SLI)
AdaptersFailed only — throttling and open circuits are expected backpressure, not defects
ClientBrowser errors only

Lifecycle

Resolve or Ignore an issue from its detail page. A resolved issue that recurs regresses back to unresolved and gets a badge — but only for an occurrence newer than the moment you resolved it, so a backlog of already-queued occurrences can't falsely reopen it. Ignored issues keep counting and stay hidden. A regression can also notify through the operational-notifier seam.

A new issue is only a badge, never an alert: a fresh deploy mints many new types at once and paging on each would be noise.

The detail page carries the raw sample stack, an hourly trend, the versions the issue was first and last seen in, and a walkable list of recent occurrences — each linking into its full trace.

See also

  • Error grouping — fingerprinting, the inbox drain, retention and config.
  • Trace — where an issue's sample occurrence leads.