摘要: Linux服务器上经常遇到一些系统和应用上的问题,如何分析排查,需要利器,下面总结列表了一些常用工具、trace tool;最后也列举了最近hadoop社区在开发发展的分布式系统的trace tool。 概览: 引用linux-performance-analysis-and-tools中图片,
Linux服务器上经常遇到一些系统和应用上的问题,如何分析排查,需要利器,下面总结列表了一些常用工具、trace tool;最后也列举了最近hadoop社区在开发发展的分布式系统的trace tool。
概览:
http://www.brendangregg.com/index.html
http://www.slideshare.net/brendangregg/linux-performance-analysis-and-tools
https://github.com/brendangregg/perf-tools/
http://www.brendangregg.com/linuxperf.html
引用linux-performance-analysis-and-tools中图片,说明这些tool试用层次位置
其中提到了的工具,大部分在我日常工具箱里或者在实践的案例里面使用过, 都有很高的价值,这里方便大家索引下:
- nicstat: 参见 这里
- oprofile: 参见 这里
- perf: 参见 这里
- systemtap: 参见 这里
- iotop: 参见 这里
- blktrace: 参见 这里
- dstat: 参见 这里
- strace: 参见 这里
- pidstat: 参见 这里
- vmstat: 参见 这里
- slabtop: 参见 这里
- tcpdump: 参见 这里
- free: 参见 这里
- mpstat: 参见 这里
- netstat: 参见 这里
- tcprstat: 参见 这里
OS系统命令
系统信息(RHEL/Fedora)
- uname -a 或 cat /proc/version #print system information
- Linux hadoopst2.cm6 2.6.18-164.el5 #1 SMP Tue Aug 18 15:51:48 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux
- uptime
- 15:42:46 up 674 days, 6 min, 35 users, load average: 1.30, 5.97, 11.53
- cat /etc/redhat-release
- Red Hat Enterprise Linux Server release 5.4 (Tikanga)
- lsb_release
- LSB Version: :core-3.1-amd64:core-3.1-ia32:core-3.1-noarch:graphics-3.1-amd64:graphics-3.1-ia32:graphics-3.1-noarch
- cat /proc/cpuinfo
- cat /proc/meminfo
- lspci – list all PCI devices
- lsusb – list USB devices
- last, lastb – show listing of last logged in users
- lsmod — show the status of modules in the Linux Kernel
- modprobe – add and remove modules from the Linux Kernel
常用命令/工具
- ps
- To print a process tree: ps -ejH / ps axjf
- To get info about threads: ps -eLf / ps axms
- ulimit -a
- lsof – list open files, UNIX一切皆文件
- lsof -p PID
- rpm/yum
- rpm -qf FILE #文件所属rpm包
- rpm -ql RPM #rpm包含文件
- /var/log/yum.log #yum 更新包日志
- /etc/XXX #系统级程序配置目录, 如
- /etc/yum.repos.d/ yum源配置
- /var/log/XXX #日志目录, 如
- /var/log/cron #crontab日志,可以查看调度执行情况
- ntpd – Network Time Protocol (NTP) daemon,同步集群中机器时间
- squid – proxy caching server,集群WebUI的代理
系统监控
- mpstat – Report processors related statistics. 注意%sys %iowait值
- vmstat – Report virtual memory statistics
- iostat – Report Central Processing Unit (CPU) statistics and input/output statistics for devices and partitions.
- netstat – Print network connections, routing tables, interface statistics, masquerade connections, and multicast memberships
- netstat -atpn | grep PID
- ganglia – a scalable distributed monitoring system for high-performance computing systems such as clusters and Grids.
- sar/tsar – Collect, report, or save system activity information; tsar是淘宝自己改进的版本
- 定时采样(每分钟),可查历史记录(默认5分钟),可弥补ganglia显示更详细信息
- iftop – the “top” bandwidth consumers shown. iftop wiki
- iotop
- vmtouch, Portable file system cache diagnostics and control
网络相关
- telnet/nc IP PORT – 确认目标端口是否可访问,只ping通不一定端口可访问,可能防火墙等禁止
- ifconfig/ifup/ifdown – configure a network interface
- traceroute – print the route packets trace to network host
- nslookup – query Internet name servers interactively
- tcpdump – dump traffic on a network, 类似开源工具 wireshark, netsniff-ng, 更多工具比较
- lynx – a general purpose distributed information browser for the World Wide Web
- tcpcp – allows cooperating applications to pass ownership of TCP connection endpoints from one Linux host to another one.
程序/进程相关
静态信息
- ldconfig – configure dynamic linker run time bindings
- ldconfig -p | grep SO 查看so是否在link cache中
- ldd – print shared library dependencies, 查看exe或so依赖的so
- nm – list symbols from object files,可grep查找是否存在相关的symbol,是否Undefined.
- readelf – Displays information about ELF files. 可现实elf相关信息,如32/64位,适用的OS,处理器
动态信息
- gdb
- cat /proc/$PID/[cmdline|environ|limits|status|…] – 进程相关信息
- pstack – print a stack trace of a running process
- pmap – report memory map of a process
java相关
- JDK Tools and Utilities
- Java Troubleshooting Tools
- jinfo – print java process information, 如classpath,java.libary.path(jni so目录)
- jstack – print a stack trace of a running java process,可查看死锁情况
- jmap – report memory map of a java process
- jmap -histo:live 可触发full gc
- jmap -dump:live,file=$FILE 可dump heap内存,用于jhat等工具debug分析object在heap的占用情况
- jhat – Heap Dump Browser – Starts a web server on a heap dump file (eg, produced by jmap -dump), allowing the heap to be browsed.
- 起http服务,浏览器访问查看
- -J-mxXXXm ,分析大文件时需要加大heap大小
- 若有对象数据超大或内存占用过多,极有可能memory leak
- Memory Analyzer (MAT) – eclipse plugin,Java heap analyzer
- 可视化工具,但受到机器内存的限制,无法分析太大的heap dump file
- jdb – 可起服务做server,eclipse等工具远程连接调试
- jstat – Java Virtual Machine Statistics Monitoring Tool
- jstatd – Virtual Machine jstat Daemon,可配合jvisualvm
- jvisualvm – Java Virtual Machine Monitoring, Troubleshooting, and Profiling Tool;可远程连接jstatd/jmx, 可视化展示工具:演示
- jvmtop – In a top-like manner, displays JVM internal metrics (e.g. memory information) of running java processes.
- JVM performance optimization JVM开发者写的优化文章
- Overview
- Compilers
- Garbage collection
- Concurrently compacting GC
- Scalability
- HPROF – Heap Profiler: java -agentlib:hprof
Trace/Debug/Profiling工具
通用工具
- 写log,但系统在线或无法源码时
- strace – trace system calls and signals
- 示例:strace/ltrace的应用实例
- 示例:可跟踪系统调用时间,如机器cpu:%sys高的问题
% time seconds usecs/call calls errors syscall
—— ———– ———– ——— ——— —————-
67.90 3966.320849 496 7992161 3050250 futex
25.80 1507.326693 127093 11860 epoll_wait
………………..
- blktrace, generate traces of the i/o traffic on block devices
- ltrace – A library call tracer
- xtrace
- gprof – a performance analysis tool, sampling and call-graph profiling
- valgrind – an instrumentation framework for building dynamic analysis tools. automatically detect many memory management and threading bugs, and profile your programs in detail
- systemtap – a simple command line interface and scripting language for writing instrumentation for a live running kernel plus user-space applications for complex tasks that may require live analysis, programmable on-line response, and whole-system symbolic access.
- Linux版DTrace(SUN在Solaris上开发的)
- 功能强大,kernel, user-space app,cross language(java perl python ruby),build-in markers(pg mysql)
- can write and reuse simple scripts to deeply examine the activities of a live system
- Data can be extracted, filtered, and summarized quickly and safely, to enable diagnoses of complex performance or functional problems
- 丰富的 “tapset” script library
java trace工具
- btrace – dynamic tracing tool for the Java platform. UserGuide
- 基于动态字节码修改技术(Hotswap)来实现运行时java程序的跟踪和替换, 实现原理
- BTrace使用总结
- 详细介绍
- byteman – simplifies tracing and testing of Java programs. Can modify a running application without needing to stop and restart it.
- define rules specifying the side effects you want to inject 而 BTrace类java语法
Distributed Tracing Tools
- Dapper, a Large-Scale Distributed Systems Tracing Infrastructure
- x-trace, a network diagnostic tool designed to provide users and network operators with better visibility into increasingly complex Internet applications.
- HTrace, a tracing framework intended for use with distributed systems written in java
- Add Tracing to HDFS
- Update HTrace for HBase
Linux observability tools | Linux 性能观测工具
Linux benchmarking tools | Linux 性能测评工具
Linux tuning tools | Linux 性能调优工具
Linux observability sar
Brendan Gregg 目前是 Netflix 的高级性能架构师 ,他在那里做大规模计算机性能设计、分析和调优。他是《Systems Performance》等技术书的作者,因在系统管理员方面的成绩,获得过 2013年 USENIX LISA 大奖。他之前是 SUN 公司是性能领头人和内核工程师,他在 SUN 开发过 ZFS L2ARC,研究存储和网络性能。他也发明和开发过一大波性能分析工具,很多已集成到操作系统中了 。他的最近工作包括研究性能分析的方法论和可视化,其目标包括Linux内核。
上面这是 Gregg 的简介,正如其中说的,他个人站点上分享了很多 Linux 性能相关的资源,都是自己开发的: