starrocks-ai-incident-rca

What problem does it solve?

This Skill helps on-call engineers quickly identify root causes behind StarRocks frontend/backend incidents by correlating symptoms like OOMs, Routine Load pauses, query timeouts, and replication or Kafka lag failures into a structured, actionable RCA.

Core Features & Use Cases

• Incident classification and evidence collection: Detects incident type (BE OOM, Routine Load pause, query timeout spike, tablet replication failure, Kafka consumer lag spike, compaction backlog) and gathers supporting data from StarRocks system views (e.g., SHOW BACKENDS, SHOW ROUTINE LOAD, tablet health signals, and query/audit log windows).
• Root cause analysis for common failure modes: Produces targeted hypotheses such as memory pressure driven by concurrent Routine Load tasks, paused job causes inferred from ErrorLogUrls/reason patterns, and recurring SQL timeout patterns via audit-log parsing.
• Structured RCA report generation: Outputs a clear incident narrative with evidence and a recommended next action suitable for operational runbooks and post-incident reviews.

Quick Start

Use the starrocks-ai-incident-rca skill to analyze a StarRocks alert by asking the agent to generate a structured RCA for a Routine Load job paused in the sales database.