Knowledge Graph Incidents

Symptoms

• SPARQL queries timing out (>5s)
• Reasoning profile returns errors
• Named graph corruption detected
• High memory usage on KG service

Immediate Mitigation

1
2
3
4
5
6
7
8
# Check KG health
curl http://localhost:4800/kg/health

# Restart KG service
just restart knowledge-graph

# Check memory usage
docker stats sea-knowledge-graph

Verification Steps

1. Health endpoint returns 200
2. SPARQL queries respond <1s
3. Reasoning profiles working
4. No error spikes in logs

Log Locations

• Service logs: docker logs sea-knowledge-graph
• OpenObserve: KG stream

Rollback

1
2
# Revert to last known good snapshot
just kg-revert-snapshot <snapshot-id>

Escalation

• P0: KG data corruption → Engage architect immediately
• P1: Query performance > SLA → Check indexing