On-Call Handoff

Transfer on-call context completely so the incoming engineer can respond effectively from minute one.

---

Why Handoffs Fail

• Ongoing incidents described as "it's fine now" when the root cause isn't fixed
• Open alerts silenced without documentation
• Workarounds applied but not written down
• Incoming engineer has to ask the same questions in the middle of the next incident
• No signal on system health heading into the shift

---

Handoff Checklist

Before your shift ends, complete this checklist:

Active Incidents

• Any P1/P2 still in progress? If yes, do NOT hand off mid-incident — stay until stable
• Any incidents resolved in the last 24h that are "resolved but watching"?
• Any open postmortem action items that affect on-call response?

Open Alerts / Silences

• Any alerts currently silenced? Why? Until when?
• Any alerts that fired but are "known flaky"? Ticket linked?
• Any Dependabot / security alerts that need immediate action?

System Health

• Error rates normal on all services? (link to dashboard)
• Any elevated latency or saturation trends?
• Any deployments in the last 24h that should be watched?
• Any scheduled jobs that failed or are overdue?

Upcoming Risks

• Any deploys scheduled in the next 24h?
• Any maintenance windows or infrastructure changes?
• Any traffic events expected (launches, marketing pushes, seasonal spikes)?

---

Handoff Document Template

## On-Call Handoff: [Date] [Your Name] → [Incoming Name]

**Shift**: [start time] → [end time]  
**Dashboard**: [link]  
**Incident tracker**: [link]  
**Runbook index**: [link]

---

### 🔴 Active Issues (needs immediate attention)

#### Issue: [Title]
- **Status**: [investigating / mitigated / monitoring]
- **Impact**: [who/what is affected]
- **Current state**: [what you know so far]
- **What to watch**: [metric, alert, log query]
- **Next action**: [what needs to happen next]
- **Ticket**: [link]

---

### 🟡 Watching (stable but monitor)

| Service | Symptom | Since | Ticket | Watch Until |
|---|---|---|---|---|
| [name] | [description] | [time] | [link] | [condition] |

---

### 🔇 Silenced Alerts

| Alert | Silenced Since | Why | Restore When |
|---|---|---|---|
| [name] | [time] | [reason] | [condition] |

---

### 📋 Recent Incidents (last 24h)

| Time | Service | Summary | Status | Postmortem |
|---|---|---|---|---|
| [time] | [service] | [1-line summary] | Resolved | [link or "in progress"] |

---

### 🚀 Upcoming Changes

| When | What | Risk | Owner |
|---|---|---|---|
| [time] | [description] | [Low/Med/High] | [name] |

---

### 💡 Tips for This Shift

[Anything unique about the current state of the system, known quirks,
 who to call for what, etc.]

**Key contacts**:
- Database issues: [name / Slack handle]
- Infra/AWS: [name / Slack handle]
- Product escalation: [name / Slack handle]

---

Synchronous Handoff (Live)

For high-stakes shifts (post-major-incident, complex active issues), do a live handoff:

Format: 15-minute sync

Agenda:

1. Walk through active issues (5 min)
2. Dashboard review — walk the incoming engineer through current state (5 min)
3. Questions from incoming engineer (5 min)

Outgoing engineer rule: Don't say "it's fine" — describe specifically why it's fine and what you'd watch for.

---

Manager: On-Call Health Signals

Watch for these signals that on-call is becoming unsustainable:

Signal	Threshold	Action
Incidents per shift	> 3	Reduce alert noise; fix recurring issues
Pages outside business hours	> 2/week avg	Audit alert thresholds and routing
"Watching" items that never resolve	> 1 week old	Prioritize fix in backlog
Same incident repeats	2+ times	Block for postmortem and systemic fix
Engineer reports burnout	1 mention	Immediate backlog reprioritization

---

Post-Shift: What to File

After your shift ends:

• Link to any new runbook entries you created
• File tickets for anything that needs fixing but wasn't urgent
• Update known-issues doc if you found new quirks
• Provide verbal/written feedback on alert quality: was anything too noisy? too silent?

---

References

• Google SRE Book: Chapter 12 — Effective Oncall
• PagerDuty: On-Call Best Practices
• LeadDev: Sustainable on-call practices

MCP Integration

Query open and in-progress incidents from Jira:

jira_search_issues(jql: "project = INC AND status in ('Open', 'In Progress') ORDER BY priority DESC, created ASC")

Read known-issue tickets linked to current incidents:

jira_get_issue(issue_key: "INC-<id>")

Search for the current runbook in Confluence:

confluence_search(query: "runbook <service_name>", space_key: "OPS", limit: 3)

Write the handoff document to Confluence:

confluence_create_page(space_key: "OPS", title: "On-Call Handoff — <Date> (<Outgoing> → <Incoming>)", content: "<formatted handoff document>", parent_page_id: "<on_call_logs_page_id>")

Graceful fallback: Output the handoff document as Markdown with a message: > 📋 Paste into Confluence > OPS > On-Call Logs > <Date>.