Capstone — Running One Incident Start to Finish

Incident Response & Ticketing Workflows

Chapter 10 · Capstone — Running One Incident Start to Finish

Nine chapters built the pieces — classification, a good ticket, a live timeline, mitigating safely, escalating well, communicating honestly, and reviewing blamelessly. Unlike this subject's other four courses, this capstone doesn't split into three separate tickets — it runs one complete, fresh incident through every one of those stages in sequence, since a process course benefits more from one coherent story than three disconnected ones.

The Incident: Search Returning Zero Results

Starting around 10:00 UTC, search begins returning no results for a growing number of previously-working queries. Browsing and checkout are unaffected.

Stage 1 — Intake (Chapter 4)

Ticket filed at 10:14 UTC: "Search returning no results for common terms since around 10am." Steps to reproduce: search for 'wireless headphones' on the site. Expected: results shown. Actual: empty results, no error visible to the user. Scope: appears to affect only some search terms, not all. Timestamp: onset ~10:00 UTC. No correlation ID yet — no visible error to capture one from, a genuine limit worth noting rather than inventing one that doesn't exist.

Stage 2 — Classification (Chapters 2 and 3)

Scoping: many users (a dozen similar reports within 15 minutes), one specific feature (search only), sudden onset, correlating with a deploy at 09:55 UTC. Severity: SEV2 — a major feature broken for a significant subset, not a full outage. Impact/Urgency: high impact (search drives significant traffic) and rising urgency (actively worsening as more terms are affected) — act now, per Chapter 3's own matrix.

Stage 3 — Live Timeline (Chapter 5)
10:16 — Confirmed several search terms return zero results; others still work. 10:19 — Checked search-index service: process healthy, resource usage normal. 10:24 — Theory: results cache serving stale/empty entries. Checked directly — cache is passing through correctly. Ruled out. 10:31 — Checked recent deploys: search-service v3.2 shipped 09:55 UTC, changed the query-parameter format sent to the index backend. 10:36 — Confirmed: v3.2 sends a terms[] array; the index backend (unchanged, still on the prior contract) silently returns empty for that format instead of erroring — matches symptom onset exactly. 10:40 — Root issue confirmed: v3.2's query-format change is incompatible with the current index backend.

Notice the 10:24 entry stays in the record — a reasonable theory, checked and ruled out, exactly as Chapter 5 requires.

Stage 4 — Mitigate vs. Fix (Chapter 6)

Decision: roll back v3.2 immediately — SEV2, actively worsening. Before rolling back, a 90-second capture: the exact query-format diff between v3.1 and v3.2, three example failing search terms with timestamps, and the index backend's own version number. A quick harm check: search is read-only, no in-flight writes at risk — the rollback itself is safe. Rolled back at 10:44 UTC.

Stage 5 — Escalation (Chapter 7)

Technical escalation, sent alongside the rollback decision: "Escalating to Search Infra — SEV2, v3.2's query-format change is incompatible with the current index backend contract (diff attached). Rolling back v3.2 as immediate mitigation, effective ~10:44 UTC. Need: a target timeline for updating the index backend to support the new format, so a compatible v3.2 re-deploy can be planned properly." No management escalation was needed this time — resolved within the hour, no business decision required. Not every incident needs both kinds, exactly as Chapter 7 notes.

Stage 6 — Communication (Chapter 8)
10:20 — Investigating reports of search returning no results for some queries. Impact: some search terms affected; browsing and checkout unaffected. Next update in 20 min. 10:40 — Identified likely cause — a recent deployment change. Preparing a rollback now. Next update within 15 min. 10:46 — Rolled back the recent deployment. Search functionality restored as of 10:44 UTC. Monitoring to confirm full resolution. 11:05 — Resolved. Search fully functional, no further reports in the last 20 minutes.
Stage 7 — Post-Incident Review (Chapter 9)
ROOT CAUSE (via 5-whys) Why did search fail? New query format returned empty results. Why? The index backend didn't support the new format. Why was it deployed anyway? No compatibility test existed between search-service and index-backend versions in the deploy pipeline. Why not? Cross-service compatibility testing was never set up for this specific pair of services. Root cause: a missing cross-service compatibility check in the deploy pipeline — not just "coordinate this one deploy better." IMPACT — SEV2, ~44 minutes (10:00–10:44 UTC), an estimated 15-20% of search queries affected during the window. WHAT WENT WELL — root cause identified in 24 minutes from ticket to confirmed cause; a clean, single-action mitigation with no side effects; a consistent communication cadence throughout. WHAT WENT POORLY — the incompatible deploy shipped with no automated check that would have caught it beforehand. ACTION ITEMS 1. Add a cross-service compatibility test between search-service and index-backend to the CI pipeline — Owner: A. Rossi — Due: 2026-08-20 2. Document the query-format contract between the two services explicitly — Owner: K. Nguyen (search-infra lead) — Due: 2026-08-18

Chapter Attribution

StageSource chapter
Ticket intake with the core actionable elementsChapter 4
Severity classification (SEV2)Chapters 2–3
Priority via the Impact/Urgency matrixChapter 3
Live timeline, including a ruled-out theory kept in the recordChapter 5
Capture-before-mitigate, plus a harm check before rolling backChapter 6
A technical escalation with evidence and a clear ask; recognizing no management escalation was neededChapter 7
A consistent, honest update cadence and an explicit all-clearChapter 8
A blameless review with a real root cause and dated, owned action itemsChapter 9
The framing distinguishing this course's own scope from the other four Technical Support coursesChapter 1

Honest Scope Note

What this course deliberately doesn't cover
  • No specific ticketing-tool tutorials (Jira, PagerDuty, and similar) — the concepts transfer directly, but tool interfaces vary too much to cover here
  • No formal ITIL-certification-level process depth — this course teaches the practical core, not a certification curriculum
  • No legal/compliance-specific incident reporting requirements (e.g. breach-notification timelines) — a genuinely separate, specialized topic
  • No crisis PR or media handling for major public-facing incidents — a distinct discipline of its own
  • No on-call rotation design or staffing methodology — this course assumes a rotation exists, not how to build one
Each is a legitimate, separate topic — not silently assumed solved by what this course actually covers.

Hands-On Exercises

Exercise 1

Explain why the 10:24 timeline entry (the cache theory, ruled out) stayed in the capstone's record instead of being removed once the real cause was found.

📄 View solution
Exercise 2

Explain why no management escalation was needed for this incident, even though it was a genuine SEV2, and what this shows about Chapter 7's own two-audience distinction.

📄 View solution
Exercise 3

Explain why the capstone's root cause is "a missing cross-service compatibility check," not "the deploy broke search," and why that distinction changed what the action items actually addressed.

📄 View solution

Chapter 10 Quick Reference — Course Complete

  • One incident, every stage: classification → ticket → timeline → mitigate-vs-fix → escalation → communication → review
  • A ruled-out cache theory stayed in the timeline; the real cause (an incompatible deploy) was found 24 minutes after the ticket landed
  • A safe, evidence-preserving rollback resolved the incident in 44 minutes; no management escalation was actually needed
  • The review's root cause was a missing CI check, not "a bad deploy" — producing action items that prevent a recurrence, not just today's incident
  • This closes Incident Response & Ticketing Workflows, 10/10 chapters — the fifth complete course under the Technical Support subject, alongside Logging & Log Analysis, Network Troubleshooting, System Monitoring & Performance Diagnosis, and Web & Application Troubleshooting