Build Error Reporting notification configurations with frequency thresholds and issue tracker integrations.
Build Error Reporting notification configurations with new error alerts, frequency thresholds, and issue tracker integrations.
Required Fields
projectIdnotificationSettings.enabledOutput will appear here...Build an Error Reporting notification configuration covering new-error alerts, frequency-based burst thresholds, per-service grouping, auto-resolution policy, and issue tracker integration. The newErrorNotification threshold controls how many occurrences trigger the first alert (commonly set to 1, so a brand-new error type pages immediately on first sighting), while frequencyNotification's threshold and windowDuration together define a rate-based alert for an existing, already-seen error suddenly spiking, two structurally different alert conditions that are easy to conflate when only one of them is actually configured.
newErrorNotification fires based on a first-occurrence threshold for an error group Error Reporting hasn't seen before, it's about novelty. frequencyNotification fires when an existing error group's occurrence rate crosses a threshold within a rolling windowDuration, it's about a known error suddenly spiking. A config with only one of the two leaves a real gap, either new errors go unnoticed until they've recurred a while, or a spike in a known error type goes unflagged.
No, autoResolveDuration marks the error group as resolved in the UI after that period of inactivity, it doesn't delete the underlying error data. If the same error recurs after being auto-resolved, it typically reopens as an active group again rather than silently continuing to accumulate under a resolved, invisible state.
Without version-level separation, an error introduced only in a new deploy gets grouped together with the same error type from the prior stable version (if it also occurs there), muddying whether the new version actually introduced a regression or whether it's a pre-existing issue. Explicit version grouping is what makes a canary-vs-stable error rate comparison meaningful.
A backend team ships a deploy that introduces a new null-pointer-style exception on an edge case affecting roughly 2% of requests, and it takes almost four hours before anyone notices, purely from a customer support ticket, because newErrorNotification.threshold had been set to 50 months earlier during a noisy period and never revisited. They use the builder to lower the threshold back to 1 for genuinely new error signatures, while adding a separate frequencyNotification rule with a higher threshold specifically for catching volume spikes in already-known errors, so future first-occurrences page immediately without also re-introducing alert fatigue from routine, already-triaged error types.
The builder requires projectId and notificationSettings.enabled to resolve, confirming notifications are switched on for the project at all, everything else (thresholds, grouping, resolution policy, issue tracker wiring) is validated as well-formed JSON, since whether your threshold and window values match your team's actual tolerance for noise versus missed signals isn't something a schema check can determine.
A threshold of 1 on newErrorNotification is aggressive but usually correct for production services, anything higher means a genuinely new failure mode has to occur multiple times silently before anyone is told, which defeats the point of fast incident detection.
mutePolicy under resolutionPolicy is a common source of 'why didn't I get paged' confusion, a muted error group stops notifying for its configured duration even while actively occurring, verify mute state specifically when debugging a missing alert for a known-recurring error.
Issue tracker integration creating a ticket per new error group without any deduplication window can flood a tracker with near-duplicate tickets during a bad deploy, pair it with sensible grouping so one root cause doesn't fan out into a dozen tickets.
Was this tool helpful?
Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.