| name | systematic-multi-service-check |
| description | After finding one root cause service, systematically check all related services for additional faults before concluding. |
| tags | ["rca","insufficient_evidence","premature_conclusion","multi_service","fault_isolation"] |
| trigger_patterns | ["When investigating a microservice incident","After identifying one failing service","When multiple services are involved in a failure","When symptoms suggest cascading or co-occurring faults"] |
| type | skill |
| confidence | evolved |
| version | 1 |
| evidence | {"train_cases":6,"pattern_frequency":6} |
Systematic Multi-Service Check
When investigating a microservice incident, never stop after finding the first failing service. Incidents frequently involve multiple services failing simultaneously from different fault types (e.g., JVMRuntimeMutator on one service, NetworkLoss/NetworkPartition on another).
Required Investigation Protocol
-
Start with zero assumptions. Always query metrics, logs, and service topology before concluding anything. Never skip investigative tool calls.
-
When you find a failing service, ask: "What other services could also be failing?" Specifically:
- Check all upstream and downstream dependencies of the failing service.
- Check all services in the same request path or call chain.
- Check services that share infrastructure (same host, same namespace, same cluster).
-
For each candidate service, check for DIFFERENT fault types. The pattern across all cases is: Service A has a compute/memory fault (JVMRuntimeMutator, CPUStress, MemoryStress) while Service B has a network fault (NetworkLoss, NetworkPartition, NetworkCorrupt). These are independent — finding one does not rule out the other.
-
Produce a complete fault matrix. Before concluding, compile a list of all services with anomalies, noting the specific fault type for each. Only submit your conclusion when you have checked every service in the affected call chain.
-
If you have zero tool calls in your trajectory, you have failed. Every RCA must include at least 2-3 investigative queries to different data sources before a conclusion is valid.