TriFleetRCA: On-Premise LLM Root Cause Analysis for Kubernetes

Kubernetes incidents are diagnosed reliably only when a root-cause system's reported gains come from incident evidence rather than scenario-specific shortcuts. We present Graph Traversal Agent, a graph-guided RCA agent that combines LLM reasoning with specialized tools. The model reasons over a typed evidence graph, while deterministic graph and tool operations collect evidence, bound the search, and check proposed verdicts. We map operational constraints, including read-only evidence collection, propagation-aware diagnosis, bounded execution, and independently validated verdicts, to a typed incident graph, a LangGraph traversal state machine, and a separate validation stage. On ITBench snapshots scored by one fixed qwen-plus judge, the audited system raises root-cause-entity F1 over an earlier iteration of the same system from 0.6087 to 0.9130 on a 23-scenario common subset. A prompt-level ablation separates prompt-tuned gains from gains that survive once scenario-specific hints are removed: the stripped-prompt configuration retains 0.6958 F1 on a 19-scenario subset. The surviving gain concentrates on ChaosMesh scenarios whose ground-truth root cause is the injected fault object already present in the evidence graph, so we report it as benchmark-coupled rather than broad cross-cluster RCA evidence. Lightweight checks, including same-judge comparison, prompt-level ablation, cascade-source checking, and a telemetry no-leak test, mark claims as supported, pending, or out of scope. We scope the work to ITBench OpenTelemetry-demo snapshots. Live-cluster trials served as an engineering stress test, but alert state and trace availability did not stay stable enough for controlled scoring, so we make no production-readiness or mean-time-to-repair claim.

Cloud-hosted large language models (LLMs) are increasingly used for root cause analysis (RCA) in AIOps pipelines, but they introduce data privacy risk, network latency, and per-query cost that scale poorly with production log volumes. We present TriCalRAG, a benchmark evaluating open-weight LLMs served locally via vLLM on a single high-memory workstation GPU (NVIDIA RTX PRO 6000, 96GB) against a classical LSTM-based log anomaly detector (DeepLog), across four real, publicly available log datasets (BGL, HDFS, Thunderbird, OpenStack). We evaluate two open-weight models (Qwen2.5-14B, Mistral-Small) under three prompting strategies: zero-shot, few-shot, and retrieval-augmented generation (RAG) over a labeled incident history, reporting accuracy, precision/recall, and F1 with bootstrap 95% confidence intervals across 3 random seeds, alongside throughput and VRAM footprint. Our results show that RAG not only improves mean F1 by 0.10-0.27 over zero-shot prompting but, more importantly, substantially stabilizes model calibration: zero-shot prompting drives both models toward near-degenerate behavior (predicting "anomaly" on up to 100% of incidents on some datasets), while RAG keeps predicted-positive rates close to the true class balance in the majority of configurations. Mistral-Small achieves higher macro-averaged F1 than Qwen2.5-14B (0.644 vs. 0.560) but exhibits calibration failures in more configurations (7 vs. 5 of 12), while running at roughly half the throughput - indicating the better model choice depends on whether a deployment prioritizes peak accuracy or predictable behavior across prompting conditions. Ablations show batching scales throughput 41 times on a single card and that 4-bit quantization reduces latency 20% with no measurable accuracy loss. We release our benchmark harness, dataset splits, and evaluation code to support reproducible on-premise AIOps research.

Identifying root causes in production microservice failures requires reasoning over large-scale, multimodal telemetry spanning metrics, logs, and traces, a problem that has proved resistant to both classical and LLM-based approaches. The OpenRCA dataset exemplifies these challenges: it is large-scale, multimodal, and lacks detailed domain knowledge, and yields 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, and 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 where failures originate, we introduce a reverse reasoning agent that, given the correct answer, identifies which signals in the extracted anomalies support it 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: the bottleneck is not data access but the agent's ability to reason over it correctly. 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 when evidence extraction is perfect: scaffold engineering and better data pipelines alone cannot close this gap; progress requires improvements at the model level.