Deep Observe: Building AI Assisted Observability forConsensus Protocols

Distributed databases and blockchain systems rely on consensus protocols to maintain consistency, yet few tools exist to observe protocol behavior in live deployments. Existing approaches typically rely on static simulation, offline trace replay, or intrusive instrumentation—each of which is costly and falls short of capturing runtime execution. DeepObserve addresses this gap as an observability framework for Apache ResilientDB. As part of the DeepObserve project, we present two eBPF-based applications—ResView and ResLens—that instrument running systems with minimal code changes. Each runs as a sidecar, decoupling telemetry from the database execution path and providing low-code, language-agnostic instrumentation. DeepObserve also exposes these tools through MCP servers, enabling AI agents to connect, query runtime behavior, visualize subprotocol execution, and inspect call-stack flamegraphs. Blog: https://blog.resilientdb.com/2026/05/... Slides: https://expolab.org/papers/deepobserv... Apache ResilientDB Resources: https://resilientdb.com/ https://expolab.resilientdb.com/ https://resai-hub.resilientdb.com/ https://blog.resilientdb.com/ https://expolab.org/ Presented by: Harish Gokul, MS CS University of California, Davis