raid10制作与效果测试

环境: CentOS release 6.6 (Final)     vmware虚拟机

1.制作raid10

[root@kevin ~]# ll /dev/sd*

brw-rw---- 1 root disk 8,  0 Dec 12 21:16 /dev/sda

brw-rw---- 1 root disk 8,  1 Dec 12 21:16 /dev/sda1

brw-rw---- 1 root disk 8,  2 Dec 12 21:16 /dev/sda2

brw-rw---- 1 root disk 8,  3 Dec 12 21:16 /dev/sda3

brw-rw---- 1 root disk 8, 16 Dec 12 21:16 /dev/sdb

brw-rw---- 1 root disk 8, 32 Dec 12 21:16 /dev/sdc

brw-rw---- 1 root disk 8, 48 Dec 12 21:16 /dev/sdd

brw-rw---- 1 root disk 8, 64 Dec 12 21:16 /dev/sde

brw-rw---- 1 root disk 8, 80 Dec 12 21:16 /dev/sdf          --准备了b,c,d,e,f五块干净的盘。

[root@kevin ~]# mdadm -C mda -l1 -n2 /dev/sd{b,c}           --将b,c制作成raid1

mdadm: Note: this array has metadata at the start and

    may not be suitable as a boot device.  If you plan to

    store '/boot' on this device please ensure that

    your boot-loader understands md/v1.x metadata, or use

    --metadata=0.90

Continue creating array? y

mdadm: Defaulting to version 1.2 metadata

mdadm: array /dev/md/mda started.

[root@kevin ~]# mdadm -C mdb -l1 -n2 /dev/sd{d,e}           --将d,e制作成raid1

mdadm: Note: this array has metadata at the start and

    may not be suitable as a boot device.  If you plan to

    store '/boot' on this device please ensure that

    your boot-loader understands md/v1.x metadata, or use

    --metadata=0.90

Continue creating array? y

mdadm: Defaulting to version 1.2 metadata

mdadm: array /dev/md/mdb started.

[root@kevin ~]# ll /dev/md

total 4

lrwxrwxrwx 1 root root   8 Dec 12 21:19 mda -> ../md127

lrwxrwxrwx 1 root root   8 Dec 12 21:19 mdb -> ../md126

-rw------- 1 root root 116 Dec 12 21:19 md-device-map

[root@kevin ~]# mdadm -C md10 -l0 -n 2 /dev/md12{6,7}       --将上边制作出来的两个raid1，制作成raid10

mdadm: Defaulting to version 1.2 metadata

mdadm: array /dev/md/md10 started.

[root@kevin ~]#

参数解释：

      -l,  --level

      -n, --raid-devices

      -C, --create

 2.分区

如果系统没有parted命令，用yum安装一下

[root@kevin ~]#  parted /dev/md/md1010

-bash: parted: command not found

[root@kevin ~]# yum provides parted

......

[root@kevin ~]# yum -y install parted         --安装分区工具

......

分区：

[root@kevin ~]# parted /dev/md/md10

GNU Parted 2.1

Using /dev/md125

Welcome to GNU Parted! Type 'help' to view a list of commands.

(parted) mklabel gpt

(parted) print

Model: Unknown (unknown)

Disk /dev/md125: 42.9GB

Sector size (logical/physical): 512B/512B

Partition Table: gpt

Number  Start  End  Size  File system  Name  Flags

(parted) mkpart primary 1M 1023M

(parted) mkpart primary 1024M 2047M

(parted) print

Model: Unknown (unknown)

Disk /dev/md125: 42.9GB

Sector size (logical/physical): 512B/512B

Partition Table: gpt

Number  Start   End     Size    File system  Name     Flags

 1      1049kB  1023MB  1022MB               primary

 2      1024MB  2047MB  1022MB               primary

(parted) quit

Information: You may need to update /etc/fstab.

[root@kevin ~]#

3.创建文件系统（格式化）

[root@kevin ~]# partx -a /dev/md/md10

BLKPG: Device or resource busy

error adding partition 1

BLKPG: Device or resource busy

error adding partition 2

[root@kevin ~]# ls /dev/md/md1*

/dev/md/md10  /dev/md/md10p1  /dev/md/md10p2

[root@kevin ~]# mkfs.ext4 /dev/md/md10p1

mke2fs 1.41.12 (17-May-2010)

Filesystem label=

OS type: Linux

Block size=4096 (log=2)

Fragment size=4096 (log=2)

Stride=128 blocks, Stripe width=256 blocks

62464 inodes, 249600 blocks

12480 blocks (5.00%) reserved for the super user

First data block=0

Maximum filesystem blocks=255852544

8 block groups

32768 blocks per group, 32768 fragments per group

7808 inodes per group

Superblock backups stored on blocks:

    32768, 98304, 163840, 229376

Writing inode tables: done

Creating journal (4096 blocks): done

Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 31 mounts or

180 days, whichever comes first.  Use tune2fs -c or -i to override.

[root@kevin ~]# mkfs.ext4 /dev/md/md10p2

......

[root@kevin ~]# mkdir /mnt/p{1,2}

4.绑定分区uuid

[root@kevin ~]# blkid |grep md125p|awk -F" " '{print $2}' |xargs -I {} echo {} >> /etc/fstab

[root@kevin ~]# vim /etc/fstab

[root@kevin ~]# mount -a

[root@kevin ~]# tail -2 /etc/fstab

UUID=015e68a9-4f77-4ba4-bfd0-74a3fa05f940  /mnt/p1 ext4 defaults,sync 0 0

UUID=aa6c0aee-557d-470d-80ad-942c43b5704e  /mnt/p2 ext4 defaults,sync 0 0

[root@kevin ~]# df -ThP

Filesystem     Type   Size  Used Avail Use% Mounted on

/dev/sda3      ext4    87G  3.5G   79G   5% /

tmpfs          tmpfs  1.9G   72K  1.9G   1% /dev/shm

/dev/sda1      ext4   488M   31M  432M   7% /boot

/dev/md125p1   ext4   944M  1.2M  894M   1% /mnt/p1

/dev/md125p2   ext4   944M  1.2M  894M   1% /mnt/p2

[root@kevin ~]#

5.测试效果

格式化sdf并挂载

[root@kevin ~]# mkfs.ext4 /dev/sdf

mke2fs 1.41.12 (17-May-2010)

/dev/sdf is entire device, not just one partition!

Proceed anyway? (y,n) y

Filesystem label=

OS type: Linux

Block size=4096 (log=2)

Fragment size=4096 (log=2)

Stride=0 blocks, Stripe width=0 blocks

1310720 inodes, 5242880 blocks

262144 blocks (5.00%) reserved for the super user

First data block=0

Maximum filesystem blocks=4294967296

160 block groups

32768 blocks per group, 32768 fragments per group

8192 inodes per group

Superblock backups stored on blocks:

    32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,

    4096000

Writing inode tables: done

Creating journal (32768 blocks): done

Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 27 mounts or

180 days, whichever comes first.  Use tune2fs -c or -i to override.

[root@kevin ~]# mkdir /mnt/sdf

[root@kevin ~]# blkid |grep sdf|awk -F" " '{print $2}' |xargs -I {} echo {} >> /etc/fstab

[root@kevin ~]# vim /etc/fstab

[root@kevin ~]# mount -a

[root@kevin ~]# tail -3 /etc/fstab

UUID=015e68a9-4f77-4ba4-bfd0-74a3fa05f940  /mnt/p1 ext4 defaults,sync 0 0

UUID=aa6c0aee-557d-470d-80ad-942c43b5704e  /mnt/p2 ext4 defaults,sync 0 0

UUID=a4a8a86b-e366-4245-9497-cd9058e6a98d  /mnt/sdf ext4 defaults,sync 0 0

[root@kevin ~]#

模拟拷贝1g数据进入不同分区，查看拷贝情况。

[root@kevin dev]# dd if=/dev/zero of=kevin.test bs=1M count=500

500+0 records in

500+0 records out

524288000 bytes (524 MB) copied, 0.384516 s, 1.4 GB/s

[root@kevin dev]# time cp -rf kevin.test /mnt/p1/

real    0m24.015s

user    0m0.021s

sys    0m1.766s

[root@kevin dev]# time cp -rf kevin.test /mnt/p2/

real    0m26.055s

user    0m0.002s

sys    0m2.741s

[root@kevin dev]# time cp -rf kevin.test /mnt/sdf/

real    0m13.752s

user    0m0.004s

sys    0m5.044s

[root@kevin dev]# time cp -rf kevin.test /tmp/

real    0m2.444s

user    0m0.002s

sys    0m0.944s

[root@kevin dev]# time sync

real    0m9.809s

user    0m0.000s

sys    0m0.001s

[root@kevin dev]#

分析这四个时间，发现与预期相悖，这是必然的，这里只是给出测试的方法。

分析：

     1.拷贝该测试文件到/mnt/p1,/mntp2时间基本一致，相同raid10上的写入速度相对固定。

     2.拷贝该测试文件到/mnt/sdf 中发现耗时约等于raid10阵列耗时的一半.这个预期的相悖。实际这个结果是正确的。

          我的实验环境是自己的笔记本，所以性能永远受限与该物理硬盘的性能。本身虚拟机的raid0条带集对速度的提升就不会很明显，而且raid1镜像集还会有反向影响。

          从物理角度分析，单盘肯定比raid10耗时少，实验总约为一半。也符合实际，因为从物理写来看，单盘写入了500M，而raid10写入了1000M。

     3.还有最后一组时间，从当前位置拷贝至/tmp下。应为linux操作系统默认磁盘写是不同步的，所以我很快记录了sync命令耗时。两个时间相加约为12.2s,也非常接近单盘500M写入磁盘的时间。