alerting

Alerting

Generate Sentry alert rule configurations and ensure observability coverage across all projects.

Prerequisites

• Sentry MCP: list_issues (to analyze current error patterns)

Standard Alert Rules

Every Sentry project should have these baseline rules:

1. New Error Alert

• Type: Issue alert
• Conditions: New issue created with level: error or fatal
• Actions: Notify #alerts Slack channel
• Frequency: Once per issue per hour
• Rationale: Catches regressions and newly introduced errors

2. Error Spike Alert

• Type: Issue alert
• Conditions: Issue seen > 50 times in 5 minutes
• Actions: Notify #alerts-critical + PagerDuty
• Frequency: Once per issue per 30 minutes
• Rationale: Detects error storms that indicate a major incident

3. Regression Alert

• Type: Issue alert
• Conditions: Previously resolved issue regressed
• Actions: Notify + assign to the person who last resolved it
• Frequency: Once per issue
• Rationale: Catches bugs that were fixed but have returned

4. High Error Rate Alert

• Type: Metric alert
• Conditions: Error count > 100 in 10 minutes
• Actions: Notify #alerts-critical
• Frequency: Every 10 minutes while condition is active
• Rationale: Aggregate health threshold — catches issues that individually are below spike threshold but collectively indicate problems

Dynamic Alert Rules

For high-frequency existing issues, generate targeted monitoring rules:

Frequency Heuristic

1. Use Sentry MCP list_issues sorted by freq (limit 10) for the project
2. For each issue with count > 50:
   • Set alert threshold at 20% of current count per hour
   • Example: issue has 500 events → alert at 100 events/hour
3. Create an issue alert for each:
   • Name: [project] Monitor: {issue title (first 50 chars)}
   • Conditions: Issue {title} > {threshold} times per hour
   • Actions: Notify #alerts
   • Frequency: Every 1 hour while active
   • Rationale: Currently #{rank} most frequent ({count} events)

Notification Routing

Severity	Channel	Escalation
error (new)	#alerts Slack	None — informational
error (spike)	#alerts-critical Slack + PagerDuty	On-call engineer
fatal	#alerts-critical Slack + PagerDuty	On-call engineer + team lead
regression	#alerts Slack + assign to last resolver	Auto-assigned

Setup Instructions

1. Go to Sentry → Project Settings → Alerts
2. Create each rule with the conditions specified above
3. Connect Slack integration for channel notifications
4. Connect PagerDuty integration for critical alerts
5. Test each rule with a synthetic error to verify routing

Coverage Audit

To audit alert coverage for a project:

1. List all current alert rules in Sentry
2. Compare against the 4 standard rules above
3. Check if high-frequency issues have targeted monitoring
4. Verify notification routing is configured correctly