存储性能测试参考手册
目录1性能测试环境检查CHECKLIST12测试过程观察项12.1观察cache使用情况12.2观察重建时间22.3观察内存使用情况22.4观察CPU使用情况22.5观察网口使用情况32021-08-03第7页, 共8页1 性能测试环境检查checklist检查项检查结果CPU配置内存配置分配的读写缓存大小磁盘型号、大小、区分监控硬盘和企业硬盘磁盘是否有坏块替换记录磁盘smart信息是否正常业务网口是否是千兆业务网口做聚合，交换机上也要配置聚合存储版本信息IMOS平台版本信息2 测试过程观察项2.1 观察cache使用情况脚本定时收集cache使用情况在日志中搜索是否有写cache满的打印The write cache of local is full.这个是每隔5min检查一次，如果检查的时候cache满了就会打印Nov 20 00:32:59 VX3000-DOWN kernel: CACHE_IO_INFO 1Total: share 0, reserve 0. LUN id: 0, blocks: dynamic 10607, actual 10607, share_used 10607, dirty 16318, send: r0 w32, flush: type 6, num 32, locus 1, wakeup 0.Nov 20 08:52:59 VX3000-DOWN kernel: CACHE_IO_INFO 1Total: share 0, reserve 0. LUN id: 0, blocks: dynamic 10607, actual 10607, share_used 10607, dirty 16318, send: r0 w32, flush: type 6, num 32, locus 1, wakeup 0.这个其实也是代表当时cache满了，接收到写IO的时候，如果未命中会去获取空闲的缓存块，如果获取不到就会打印上面信息！隔700ms就会去重试一次，5个700ms后，就透写。蓝色的数字是代表第几个700ms2.2 使用丢包工具查看是否有丢包telnet到EC或者IPC上，将丢包工具使用ftp上传然后按如下步骤执行命令：# fdisk -lDisk /dev/sda: 74.0 GB, 74088185856 bytes255 heads, 63 sectors/track, 9007 cylindersUnits = cylinders of 16065 * 512 = 8225280 bytesDisk /dev/sda doesnt contain a valid partition table# disktool -d /dev/sda super= super data info = Magic: 1587, EcPortId: HIC3401-125_1 Start storage time: year = 2013 mon = 02 day = 16 hour = 07 min = 36 sec = 59 Last wirte super data time: year = 2013 mon = 02 day = 19 hour = 14 min = 02 sec = 16 # # # disktool -d /dev/sda ui-data 0 /这步看一级索引，看出问题的时间在哪一段= unit data 0 info = 1: year = 2013 mon = 02 day = 16 hour = 20 min = 48 sec = 17 2: year = 2013 mon = 02 day = 16 hour = 21 min = 05 sec = 31 3: year = 2013 mon = 02 day = 16 hour = 21 min = 22 sec = 50 4: year = 2013 mon = 02 day = 16 hour = 21 min = 40 sec = 08 5: year = 2013 mon = 02 day = 16 hour = 21 min = 57 sec = 28 # disktool -d /dev/sda di 1 5 /这步查二级索引，比如看1-5时间段的信息，执行后会生成一个log文件。= data index head info = Display 1th unit Current writeing 1034th items di->items = 1035 Write position 264323072 di->pos = 264323072= data index head info = data index head info = Display 2th unit Current writeing 1039th items di->items = 1040 Write position 264454144 di->pos = 264454144= data index head info = data index head info = Display 3th unit Current writeing 1038th items di->items = 1039 Write position 264372224 di->pos = 264372224= data index head info = data index head info = Display 4th unit Current writeing 1039th items di->items = 1040 Write position 264470528 di->pos = 264470528= data index head info = data index head info = Display 5th unit Current writeing 1033th items di->items = 1034 Write position 264306688 di->pos = 264306688= data index head info =dump data_index to ./di_sda_1_to_5.logdi_start = 1, di_end = 5# more di_sda_1_to_5.log /打开生成的log就可以了，看哪一秒丢数据了，哪一条就是没有记录= data index head info = Display 1th unit Current writing 1034th items di->items = 1035 Write position 264323072 di->pos = 264323072= data index head info = 1: 2013-02-16, 20:48:17, reserved: 0, blocks = 15, offset = 16 2: 2013-02-16, 20:48:18, reserved: 0, blocks = 15, offset = 31 3: 2013-02-16, 20:48:19, reserved: 0, blocks = 16, offset = 46 4: 2013-02-16, 20:48:20, reserved: 0, blocks = 15, offset = 62 5: 2013-02-16, 20:48:21, reserved: 0, blocks = 15, offset = 77 6: 2013-02-16, 20:48:22, reserved: 0, blocks = 16, offset = 92 7: 2013-02-16, 20:48:23, reserved: 0, blocks = 16, offset = 108 8: 2013-02-16, 20:48:24, reserved: 0, blocks = 15, offset = 124 9: 2013-02-16, 20:48:25, reserved: 0, blocks = 16, offset = 139可使用下面的工具查看丢包情况 # ./eclosscheck.sh sda 338 350total lost on sda-338: 25 / 1057total lost on sda-339: 0 / 1047total lost on sda-340: 0 / 1061total lost on sda-341: 14 / 1074total lost on sda-342: 0 / 1061total lost on sda-343: 0 / 1061total lost on sda-344: 0 / 1060total lost on sda-345: 25 / 1074如果发现EC或者IPC有丢包，首先看写cache是否满过，如果cache没有满就需要检查网络原因导致丢包。2.3 观察重建时间脚本定时收集重建进度，估算重建时间。如果重建阵列的RAID LUN分配到的写cache的share有时会用完的话，重建会sleep，这种情况下GUI上显示的重建剩余时间会有比较大的变化，不能以GUI上显示的剩余时间来估算重建时间，需要根据脚本收集的重建进度来估算，建议5min收集一次重建进度。2.4 观察回放观察回放时
