lsf_job_slow

What problem does it solve?

This Skill helps you diagnose why an LSF/OpenLava/Volclava job stays in RUN状态 but shows no output, stalls, or progresses far slower than expected, so you can distinguish application blocking from node, system, or network bottlenecks.

Core Features & Use Cases

• Collect job and process signals: Uses LSF commands to pull execution host, runtime, CPU time, memory limits/usage, working directory, and the latest stdout/stderr evidence.
• Diagnose execution-node pressure and IO bottlenecks: Checks node load (CPU/memory/I-O/tmp space indicators) to identify overload, memory exhaustion/swap, sustained IO wait, or disk-full symptoms.
• Classify process state and likely root cause: Interprets Linux process stat (especially D-state for uninterruptible IO sleep) to pinpoint NFS/network storage hangs or disk IO failures.
• Summarize evidence into actionable recommendations: Produces a structured report with likely root causes, immediate mitigation, and prevention guidance.

Use Case: A batch job appears RUN for hours with no new stdout/stderr, and you need to quickly determine whether it is blocked on NFS, starved by node overload, waiting for license, deadlocked, or simply slow.

Quick Start

Ask the AI to diagnose the stuck job by telling it the job id you want to investigate.