Linux服务器性能排查实战手册
Linux Server Performance Troubleshooting Handbook
| iDev Tech | 2026-07-12T10:00:00
系统化的Linux服务器性能排查方法论，涵盖CPU、内存、磁盘IO和网络四大维度。
A systematic methodology for Linux server performance troubleshooting covering CPU, memory, disk I/O, and networking.
排查方法论：USE方法Brendan Gregg提出的USE方法是性能排查的黄金框架：对每种资源（CPU、内存、磁盘、网络），检查Utilization（使用率）、Saturation（饱和度）和Errors（错误数）。快速定位：uptime（负载）→ dmesg（内核错误）→ vmstat（系统概览）→ 深入排查。CPU排查top/htop查看进程CPU占用；perf top分析热点函数；mpstat -P ALL查看各核心负载均衡；判断是usr（应用代码）还是sys（系统调用）消耗。常见问题：死循环、锁竞争、GC频繁。内存排查free -h查看内存使用；/proc/meminfo获取详细信息；valgrind检测内存泄漏。注意区分buffer/cache和实际已用内存。OOM Killer触发时检查/var/log/messages。磁盘IO：iostat -xz 1查看IOPS和延迟；iotop定位IO密集进程。
Methodology: USE MethodBrendan Gregg's USE method is the gold standard for performance troubleshooting: for each resource (CPU, memory, disk, network), check Utilization, Saturation, and Errors. Quick triage: uptime (load) → dmesg (kernel errors) → vmstat (system overview) → deep dive.CPU Troubleshootingtop/htop for process CPU usage; perf top for hotspot function analysis; mpstat -P ALL for per-core load balance; determine if consumption is usr (application code) or sys (system calls). Common issues: infinite loops, lock contention, frequent GC.Memory Troubleshootingfree -h for memory usage overview; /proc/meminfo for detailed info; valgrind for leak detection. Distinguish between buffer/cache and actual used memory. Check /var/log/messages when OOM Killer triggers. Disk I/O: iostat -xz 1 for IOPS and latency; iotop to locate I/O-intensive processes.