对于我们日常的维护和分析来说，保存一份主机日常性能采集的数据是非常有必要的。可以让我们在事后查看需要查看的那段时间的主机资源使用情况，也可以据此，我们来做一份分析报告。

1、AIX平台在AIX平台下，我们可以借助nmon工具很方便的实现这个需求。我们需要做的就是上传nmon工具，然后在crontab中部署一个计划任务。0 0 * * * cd /home/xigua ; ./nmon_aix53 -x >> nmon.log 2>&1-x代表每隔15分钟做一次采样，当然如果你认为这个频率太低，可以修改。例如：0 0 * * * nmon64 -ft -D -E -I 1 -s 120 -c 720 -m /aixpatch/nmon_out改成2分钟采集一次，每天采集720次。2、HP-UNIX平台在hp-unix下面比较麻烦一点，我们需要自己来写采集的脚本。脚本主要是调用glance来采集数据。定义的crontab如下：0 0 * * * cd /glance; ./glance.sh/glance# more glance.sh## This scripts use glance plus to monitor the performance .####!/sbin/shHOST=`hostname`DAY=`date '+%Y%m%d'`cd /glance/# NETWORK INTERFACE LOOPnohup /opt/perf/bin/glance -adviser_only -nosort -iterations 288 -j 300 -syntax ./net.adviser >>./log/net_${HOST}_${DAY}.log &# SYSTEM LOOPnohup /opt/perf/bin/glance -adviser_only -nosort -iterations 288 -j 300 -syntax ./sys.adviser >>./log/sys_${HOST}_${DAY}.log &#PROCESS LOOPnohup /opt/perf/bin/glance -adviser_only -nosort -iterations 288 -j 300 -syntax ./process.adviser >>./log/process_${HOST}_${DAY}.log &#DISK LOOPnohup /opt/perf/bin/glance -adviser_only -nosort -iterations 288 -j 300 -syntax ./disk.adviser >>./log/disk_${HOST}_${DAY}.log &这个脚本负责调用4个模块的数据采集，采集的频率是5分钟。4个模块的采集数据定义如下 ：cat sys.adviser# This adviser script. prints the global cpu/disk/mem/swap utilization and also # the network subsystem wait percent for the interval. headers_printed=headers_printedIF headers_printed == 0 THEN{ print "                       CPU    SYS   USER   DISK       PHY_READ      PHY_WRITE    MEM     SWAP  Network"print " Date      Time       Util   Util   Util   Util           RATE           RATE   Util     Util   Wait %"print "------------------------------------------------------------------------------------------------------"headers_printed=1}print GBL_STATDATE," ",GBL_STATTIME," ",          GBL_CPU_TOTAL_UTIL," ",GBL_CPU_SYS_MODE_UTIL," ",GBL_CPU_USER_MODE_UTIL," ",          GBL_DISK_UTIL_PEAK," ",GBL_DISK_PHYS_READ_BYTE_RATE,"KBs ",GBL_DISK_PHYS_WRITE_BYTE_RATE,"KBs ",          GBL_MEM_UTIL,"   ",          GBL_SWAP_SPACE_UTIL," ",          GBL_NETWORK_SUBSYSTEM_WAIT_PCT/glance# cat net.adviser# This version will only work with hp-ux 11.x.If you want it to# work for 10.20 you need to remove the "BYNETIF_QUEUE," string# below as that metric is only available from 11.x glance.# The following string variable should be changed to the interface# of interest. For example:# netif_to_examine = "lan0"# If you want to see all interfaces, leave it an empty string (""):# netif_to_examine = ""# initialize variables:# headers_printed=headers_printednetif_to_examine=""# print headers before the loopprint "                     NetIF               NetIn      NetOut        Net        Net"print "  Date     Time       Name             bytes/s     bytes/s  COLLISION      ERROR"print "----------------------------------------------------------------------------------"# begin loopnetif loop { # print information for the selected interface or if null then all:        IF (BYNETIF_NAME == netif_to_examine) or (netif_to_examine == "") THEN        {                 IF BYNETIF_IN_BYTE_RATE > 0 or BYNETIF_OUT_BYTE_RATE > 0 THEN                { # print one line per interface reported which has byte_rate > 0:                print GBL_STATDATE," ",GBL_STATTIME,"   ",BYNETIF_NAME," ",                                  BYNETIF_IN_BYTE_RATE," ",BYNETIF_OUT_BYTE_RATE," ",                                  BYNETIF_COLLISION," ",BYNETIF_ERROR                }# (note that some interface types do not report collisions or errors)        }}print ""/glance# cat disk.adviser## This adviser is used for capture the disk infomation .# # print headers before the loopprint "                        Mounted           DISK RESPONSE       PHY_READ      PHY_WRITE  DISK "print "  Date     Time      Filesystem           IOPS     TIME           RATE           RATE  Util "print "-------------------------------------------------------------------------------------------------"DISK LOOPif BYDSK_DIRNAME != "Unknown" and (BYDSK_PHYS_READ_BYTE_RATE > 0 OR BYDSK_PHYS_WRITE_BYTE_RATE > 0OR BYDSK_UTIL > 0)then { print GBL_STATDATE," ",GBL_STATTIME," ",BYDSK_DIRNAME|15|0,BYDSK_PHYS_IO_RATE,BYDSK_AVG_SERVICE_TIME,"ms ",BYDSK_PHYS_READ_BYTE_RATE,"KBs ",BYDSK_PHYS_WRITE_BYTE_RATE,"KBs",BYDSK_UTIL}/glance# cat process.adviser## This adviser is used to capture the process infomation .# # print headers before the loopprint "                                                                   PROC  PROC  "print "                      PORC     PROC       PROC       PROC   PROC    LOG   PHY        PHY_IO   PROC "print "  Date     Time      OWNER       ID        MEM   MEM_HIGH    CPU   IOPS  IOPS          RATE   NAME "print "-----------------------------------------------------------------------------------------------------"PROCESS LOOPif PROC_MEM_RES > 100000 then { print GBL_STATDATE," ",GBL_STATTIME," ",PROC_USER_NAME,PROC_PROC_ID|7|0," ",PROC_MEM_RES," ",PROC_MEM_RES_HIGH," ",PROC_CPU_TOTAL_TIME,PROC_DISK_LOGL_IO_RATE,PROC_DISK_PHYS_IO_RATE,PROC_IO_BYTE_RATE,"KBs ",PROC_PROC_NAME}

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/12330444/viewspace-584302/，如需转载，请注明出处，否则将追究法律责任。