Troubleshooting detection & notifications

Work these symptom-first. Each section states what you are seeing, why it happens, and what to check.

An expected issue was not detected

The agent opens issues from what it finds during an investigation, not from static thresholds — so undetected usually means it never looked, or the data wasn't there.

• The data isn't arriving. Open the Intake page and confirm logs from the relevant source are landing. If volume is zero, the problem is ingestion, not detection — see Troubleshooting ingestion.
• Nothing investigated that window. Detection follows investigation. If no automation or question covered the affected time and service, no issue opens. Consider a scheduled automation that samples the signal, or the Sazabi-managed Investigate new error patterns signal automation.
• The issue exists but is muted. A muted issue stays open and does not re-notify. Check the inbox for a muted match and review its mute hit count.

Several issues for one incident

You expected one issue but got several.

• Sazabi matches a recurrence to an existing open issue by its stable name. Distinct symptoms of one root cause can legitimately produce distinct issues — a database outage may surface as separate error patterns in several services.
• Link the related issues to the same component so the status page shows one degraded service rather than scattered issues.
• Resolve the duplicates you don't need; resolved issues leave the active inbox but stay searchable.

I didn't get notified

An issue opened, or an automation failed, but nothing reached you.

• Personal channel not on. Personal delivery is gated by your preference. Open Settings > Notifications and confirm the channel is on for that event type. In-app and email are on by default; Slack DM and SMS are opt-in.
• No matching delivery rule. Shared delivery is positive — a shared channel receives an event only when a rule names both the event type and that destination. Check the project Settings > Notifications page for a rule covering the event.
• A condition filtered it out. If the rule has a severity or component condition, an issue below the minimum severity or on another component won't match. Raise or lower the rule's severity condition under Settings > Notifications.
• Terminal events skip email by default. Resolved and ignored events stay in the bell feed but do not email unless you turned email on for them.

See Route notifications for the full routing model.

Getting help

If an issue still isn't detected or delivered after these checks, gather the project, the affected time window, the issue name (if one exists), and the delivery rule or personal preference you expected to match, then contact support.

Further reading