triage

Triage

Diagnose and triage Sentry issues using the official Sentry MCP server.

Prerequisites

You need the Sentry MCP server connected. The following Sentry MCP tools are used:

• list_issues — list issues by project with search syntax
• get_issue_details — full issue detail with stack trace, tags, contexts
• list_issue_events — event history for an issue
• update_issue — resolve, ignore, or unresolve an issue
• search_issues — find issues by trace ID or other criteria
• get_project_stats — error rate trends over time

Sentry Projects

The Lucid platform uses these Sentry projects:

• lucid-web — Next.js frontend (Vercel)
• lucid-worker — Worker service (Railway/Fastify)
• lucid-l2 — L2 gateway (Railway/Express)
• lucid-trustgate — Trust gate (Railway/Fastify)
• lucid-mcpgate — MCP gate (Railway/Fastify)
• javascript-nextjs — Next.js SSR errors

Sentry org: raijin-labs

Triage Procedure

Step 1: Get Issue Details

Use Sentry MCP get_issue_details with the issue ID. Focus on:

• Stack trace: Last 15 frames, prioritize frames marked [in_app]
• Tags: Look for trace_id, run_id, service, environment, release
• Breadcrumbs: Last 10 breadcrumbs for event timeline context
• OTel context: Check contexts.otel for trace correlation

Step 2: Analyze Temporal Pattern

Use Sentry MCP list_issue_events (limit 25) and classify the event timing pattern:

• See references/temporal-patterns.md for the classification algorithm
• The pattern tells you whether this is a burst (deployment?), steady (systemic), regression (reintroduced), or sporadic (edge case)

Step 3: Check Cross-Service Correlation

If the issue has trace_id or run_id tags:

• Use Sentry MCP search_issues with query trace_id:{id} across all 6 projects
• If errors appear in >1 service → cascade detected — investigate the oldest error first
• See the correlation skill for detailed cross-service procedure

Step 4: Match Diagnosis Pattern

Compare the issue title, culprit, and stack trace against known patterns:

• Check Lucid-specific patterns first (see references/diagnosis-patterns.md)
• Then check built-in patterns (network, timeout, auth, rate limit, database, validation, memory)
• Matching is case-insensitive substring match on title + culprit + stack trace text

Step 5: Check Known Bugs

Compare against references/known-bugs.md:

• If a known bug matches and status: FIXED, recommend verifying the fix is deployed
• If a known bug matches and status: OPEN, link to the existing fix

Step 6: Score Severity and Recommend Action

Use references/severity-matrix.md to score:

• CRITICAL → Immediate escalation, consider rollback
• HIGH → Link to deploy diff, suggest revert if post-deployment
• MEDIUM → Create investigation ticket
• LOW → Monitor; auto-resolve if <5 occurrences/day

Issue Resolution

When to use Sentry MCP update_issue:

• Resolve: Root cause identified AND fix deployed AND verified in production
• Ignore: Known noise (e.g., bot traffic, expected transient errors). Set duration if temporary.
• Unresolve: Previously resolved but still occurring — regression detected

Error Rate Analysis

Use Sentry MCP get_project_stats to analyze trends:

• stat: received — total errors ingested
• stat: rejected — errors dropped by rate limits
• interval: 1h for recent spikes, 1d for weekly trends
• period: 7d or 14d for baseline comparison
• Compare current rate to 7-day average — >2x is a spike, >5x is an incident