Home›🔍 Playbooks›Module 72 min read · 8/8

Davis AI & Diagnostics

Reference

Davis AI & Diagnostic Tools

Dynatrace provides AI-powered tools for troubleshooting: Davis Problems, Davis CoPilot, OneAgent Diagnostics, and Dynatrace Assist.

Davis Problems — Always Start Here

// Active problems — the FIRST query in any investigation
fetch dt.davis.problems, from:now()-2h
| filter not(dt.davis.is_duplicate) AND event.status == "ACTIVE"
| fields display_id, event.name, event.category, event.start,
        root_cause_entity_name, dt.davis.affected_users_count
| sort event.start desc

⚠️ ALWAYS start with Davis problems, never broad log searches. Broad log queries hit scan limits and return 0 results. Scope ALL queries to problem context (entity + timeframe).

Correct Problem Field Names

WRONG                  CORRECT
─────────────────────  ─────────────────────
title                  event.name
status                 event.status
severity               event.category
start                  event.start
"OPEN"                 "ACTIVE" (OPEN does not exist!)

Problem-Scoped Investigation

// Step 1: Get problem details
fetch dt.davis.problems
| filter display_id == "P-12345"
| fields event.start, event.end, event.description,
        root_cause_entity_id, root_cause_entity_name

// Step 2: Scope logs to problem timeframe (±5 min)
fetch logs, from:problemStart-5m, to:problemEnd+5m
| filter dt.entity.service == "SERVICE-XXX"
| filter loglevel == "ERROR"
| sort timestamp desc

// Step 3: Find trace IDs in logs for correlation
fetch logs, from:problemStart-5m, to:problemEnd+5m
| filter isNotNull(trace_id)
| fields trace_id, content, loglevel

Davis CoPilot (AI Assistant)

  • Natural language → DQL: type a question in Dashboards/Notebooks, get auto-generated DQL
  • DQL → Natural language: explain existing queries in plain English
  • Dynatrace Assist: contextual AI across apps — summarizes problems, explains K8s warnings, suggests troubleshooting guides

API scopes: davis-copilot:nl2dql:execute, davis-copilot:dql2nl:execute, davis-copilot:conversations:execute

OneAgent Diagnostics

Automated troubleshooting for OneAgent issues:

  1. Go to Hosts Classic → select host → ... → Run OneAgent diagnostics
  2. Dynatrace collects 24h of diagnostic data
  3. Auto-detects: NFS mounts, connectivity issues, version mismatches, deep monitoring failures
  4. Results include potential solutions with instructions

CLI alternative: oneagentctl --create-support-archive

Structured Troubleshooting Playbook

Step  Action                                    Tool
────  ──────────────────────────────────────    ──────────────────
1     List active Davis problems                 dt.davis.problems
2     Select problem → extract entities + time   Problem details
3     Scope: queryFrom = start-5min              DQL timeframe
4     Query logs for affected entities           fetch logs
5     Classify errors (actionable vs expected)   Manual analysis
6     Find trace_id → reconstruct request flow   fetch spans
7     Summarize: root cause + next steps         Report
▶ Knowledge Check

Q: What's the correct status value for active problems?

  • ❌ "OPEN"
  • ✅ "ACTIVE"
  • ❌ "IN_PROGRESS"

Q: Why should you never start with a broad log search?

  • ✅ It hits scan limits and returns 0 results
  • ❌ Logs are not stored in Grail
  • ❌ You need a special scope for logs
← PreviousKubernetes Issues