Retellai Incident Runbook

Retell AI Incident Runbook

Overview

Rapid incident response procedures for Retell AI-related outages.

Prerequisites

• Access to Retell AI dashboard and status page
• kubectl access to production cluster
• Prometheus/Grafana access
• Communication channels (Slack, PagerDuty)

Severity Levels

Quick Triage

set -euo pipefail
# 1. Check Retell AI status
curl -s https://status.retellai.com | jq

# 2. Check our integration health
curl -s https://api.yourapp.com/health | jq '.services.retellai'

# 3. Check error rate (last 5 min)
curl -s localhost:9090/api/v1/query?query=rate(retellai_errors_total[5m])  # 9090: Prometheus port

# 4. Recent error logs
kubectl logs -l app=retellai-integration --since=5m | grep -i error | tail -20

Decision Tree

Retell AI API returning errors?
├─ YES: Is status.retellai.com showing incident?
│   ├─ YES → Wait for Retell AI to resolve. Enable fallback.
│   └─ NO → Our integration issue. Check credentials, config.
└─ NO: Is our service healthy?
    ├─ YES → Likely resolved or intermittent. Monitor.
    └─ NO → Our infrastructure issue. Check pods, memory, network.

Immediate Actions by Error Type

401/403 - Authentication

set -euo pipefail
# Verify API key is set
kubectl get secret retellai-secrets -o jsonpath='{.data.api-key}' | base64 -d

# Check if key was rotated
# → Verify in Retell AI dashboard

# Remediation: Update secret and restart pods
kubectl create secret generic retellai-secrets --from-literal=api-key=NEW_KEY --dry-run=client -o yaml | kubectl apply -f -
kubectl rollout restart deployment/retellai-integration

429 - Rate Limited

set -euo pipefail
# Check rate limit headers
curl -v https://api.retellai.com 2>&1 | grep -i rate

# Enable request queuing
kubectl set env deployment/retellai-integration RATE_LIMIT_MODE=queue

# Long-term: Contact Retell AI for limit increase

500/503 - Retell AI Errors

set -euo pipefail
# Enable graceful degradation
kubectl set env deployment/retellai-integration RETELLAI_FALLBACK=true

# Notify users of degraded service
# Update status page

# Monitor Retell AI status for resolution

Communication Templates

Internal (Slack)

🔴 P1 INCIDENT: Retell AI Integration
Status: INVESTIGATING
Impact: [Describe user impact]
Current action: [What you're doing]
Next update: [Time]
Incident commander: @[name]

External (Status Page)

Retell AI Integration Issue

We're experiencing issues with our Retell AI integration.
Some users may experience [specific impact].

We're actively investigating and will provide updates.

Last updated: [timestamp]

Post-Incident

Evidence Collection

set -euo pipefail
# Generate debug bundle
./scripts/retellai-debug-bundle.sh

# Export relevant logs
kubectl logs -l app=retellai-integration --since=1h > incident-logs.txt

# Capture metrics
curl "localhost:9090/api/v1/query_range?query=retellai_errors_total&start=2h" > metrics.json  # 9090: Prometheus port

Postmortem Template

## Incident: Retell AI [Error Type]
**Date:** YYYY-MM-DD
**Duration:** X hours Y minutes
**Severity:** P[1-4]

### Summary
[1-2 sentence description]

### Timeline
- HH:MM - [Event]
- HH:MM - [Event]

### Root Cause
[Technical explanation]

### Impact
- Users affected: N
- Revenue impact: $X

### Action Items
- [ ] [Preventive measure] - Owner - Due date

Instructions

Step 1: Quick Triage

Run the triage commands to identify the issue source.

Step 2: Follow Decision Tree

Determine if the issue is Retell AI-side or internal.

Step 3: Execute Immediate Actions

Apply the appropriate remediation for the error type.

Step 4: Communicate Status

Update internal and external stakeholders.

Output

• Issue identified and categorized
• Remediation applied
• Stakeholders notified
• Evidence collected for postmortem

Error Handling

Examples

One-Line Health Check

set -euo pipefail
curl -sf https://api.yourapp.com/health | jq '.services.retellai.status' || echo "UNHEALTHY"

Resources

• Retell AI Status Page
• Retell AI Support

Next Steps

For data handling, see retellai-data-handling.