page-cache-reclaim-diagnosis

High kswapd CPU usage and system stalls on Linux are caused by kernel memory reclaim pressure. This Skill analyzes vmstat, meminfo, and zoneinfo metrics to isolate whether NUMA imbalance, memory fragmentation, or page cache overuse is triggering the reclaim storm.

You can diagnose direct memory reclaim latency by analyzing allocstall events and vmstat metrics for a specific time window. This Skill uses a hypothesis-driven framework to map process stalls to root causes like disk I/O bottlenecks or improper cache clearing, providing actionable tuning recommendations.

Dirty page writeback storms are caused by excessive dirty cache flushing that overwhelms disk I/O capacity. This Skill identifies these storms by analyzing kernel memory reclaim metrics and maps the symptoms to specific diagnostic branches to validate disk I/O bottlenecks.

To tune sysctl parameters against page cache overuse, you must first isolate the root cause of memory pressure. This Skill diagnoses kswapd spikes and direct reclaim issues using memory metrics, then provides specific sysctl tuning recommendations to mitigate cache overuse.

Yes, vmstat and meminfo analysis can be used for NUMA imbalance diagnosis. This Skill implements a structured multi-branch approach to validate potential causes like NUMA imbalance by analyzing these exact metrics alongside zoneinfo data during memory pressure events.

Diagnosing memory pressure without zoneinfo metrics limits the ability to detect memory fragmentation across zones. This Skill requires vmstat, meminfo, and zoneinfo together to validate hypotheses like NUMA imbalance and fragmentation, ensuring accurate root cause isolation.