Identifying root causes in production microservice failures requires reasoning over large-scale, multimodal telemetry spanning metrics, logs, and traces. This problem has proven resistant to both classical and LLM-based approaches. The OpenRCA dataset exemplifies these challenges: it is large-scale, multimodal, and lacks detailed domain knowledge, resulting in consistently low accuracy across all existing methods.
We show that classical causal discovery methods and existing LLM-based multi-agent systems fail to reliably identify root causes on this benchmark. To address this, we present a Structured Multi-Agent RCA pipeline that substantially outperforms existing LLM-based and classical baselines, supporting both domain-knowledge and knowledge-free operating modes.
To diagnose the origins of failures, we introduce a reverse reasoning agent that identifies which signals in the extracted anomalies support the correct answer and determines whether Stage 1 had access to those signals, classifying each failure as Reasoning Gap (evidence present but unused) or Data Ambiguity (evidence genuinely absent).
This analysis reveals that the required evidence is present in the vast majority of failures, indicating that the bottleneck is not data access but the agent's reasoning capability.
We further introduce an automated rule mining pipeline that systematically extracts discrimination rules from reverse reasoning reports, reducing reliance on manual knowledge curation. Across all configurations, model reasoning capability and domain knowledge are the primary constraints: stronger models embed more domain expertise, and explicit knowledge injection partially compensates for this gap.
Reasoning performance remains practically bounded even with perfect evidence extraction; thus, improvements at the model level are necessary for progress.
Blogger's Review: This study clearly highlights the importance of reasoning capabilities and their relation to domain knowledge in diagnosing complex system failures. By introducing a reverse reasoning agent and an automated rule mining approach, it presents new avenues to enhance root cause analysis performance, which is worth exploring in practical applications.