概述
为了满足在性能和冗余等方面的需求,LVM支持了下面三种Logic Volume:
- Linear Logic Volume --线性逻辑卷
- Striped Logic Volume --条带化逻辑卷
- Mirror Logic Volume --镜像逻辑卷
Linear Logic Volume
我们用lvcreate命令默认创建出来的就是线性逻辑卷,线性逻辑卷使用的PE可以来自一个PV,也可以来自多个PV,一般情况下是先从第一个PV中分配PE,如果这个PV的PE已经分配完了,再依次从第二个PV、第三个PV里分配。可以通过指定PV甚至PE的号段来让Linear LV的PE分散到各个PV上,但是如果其中一个PV坏了,那么这个Linear LV可能也就没法用了。Linear LV的size可以直接用-L指定大小,也可以用-l指定分配PE的个数。往Linear LV中写入数据时,先往第一个PV的PE中写,直到第一个PV上分配的空间用完了才会将数据写到第二个PV。
Linear LV只能满足弹性分配的需求,无法满足性能和冗余的需求,是最普通的volume,但是Linear LV也可以通过lvconvert命令切换成Mirror LV来提供冗余能力。
- root@hunk-virtual-machine:/home/hunk# lvcreate -l 100 -n linearlv VolGroup1 /dev/sdc:1280-1305 /dev/sdd:1280-1
- 305 /dev/sde:1280-1305 /dev/sdf:1280-1305
Striped Logic Volume
Striped LV的底层存储布局类似于RAID0,它是跨多个PV的,具体是跨多少个PV用-i指定,但是肯定不能超过VG中PV的数量,Striped LV的最大size取决于剩余PE最少的那个PV。
Striping的意思是将每个参与Striping的PV划分成等大小的chunk(也叫做stripe unit),每个PV同一位置的这些chunk共同组成一个stripe。比如下面这张图(来自于RedHat6官方文档),包含三个PV,那么红色标识的1、2、3这3个chunk就组成了stripe1,4、5、6组成stripe2。chunk的大小可以通过-I或者--stripesize来指定,但是不能超过PE的大小。
比如,向Striped LV写入数据时,数据被分成等大小的chunk,然后将这些chunk顺序写入这些PV中。这样的话就会有多个底层disk drive并发处理I/O请求,可以得到成倍的聚合I/O性能。还是下面这张图,假如现在有一个4M数据块需要写入LV,stripesize设置的512K,LVM把它切成8个chunk,分别标识为chunk1、chunk2...,这些chunk写入PV的顺序如下:
- chunk1写入PV1
- chunk2写入PV2
- chunk3写入PV3
- chunk4写入PV1
- ...
因为LVM无法判断多个Physics Volume是否来自同一个底层disk,如果Striped LV使用的多个Physics Volume实际上是同一个物理磁盘上的不同分区,就会导致一个数据块被切成多个chunk分多次发给同一个disk drive,这种情况实际上Striped LV并不能提升性能,反而会使性能下降。所以说,Striped LV提升I/O性能的本质是让多个底层disk drive并行处理I/O请求,而不是表面上的把I/O分散到了多个PV上。
Striped LV主要满足性能需求,没有做任何冗余,所以没有容错能力,如果单个disk损坏,就会导致数据损坏。
root@hunk-virtual-machine:/home# lvcreate -L 20G --stripes 4 --stripesize 256 --name stripevol VolGroup1
Mirror Logic Volume
Mirror LV就是各个PV之间做冗余,类似于RAID1,通过-m指定冗余数量。Mirror LV提供冗余能力,可以有效解决磁盘单点故障问题,但是性能方面没有帮助。Linear LV和Mirror LV直接用lvconvert工具来相互切换,Mirror LV在创建后也可以更改冗余数,具体用法请参考man page。
- root@hunk-virtual-machine:/home/hunk# lvcreate -l 100 -m1 -n mirrorvol VolGroup1
- Logical volume "mirrorvol" created.
- root@hunk-virtual-machine:/home/hunk# lvdisplay /dev/VolGroup1/mirrorvol -m
- --- Logical volume ---
- LV Path /dev/VolGroup1/mirrorvol
- LV Name mirrorvol
- VG Name VolGroup1
- LV UUID YxgfYi-c7nK-wk4v-rlu1-vRdh-MTMb-uVfl2v
- LV Write Access read/write
- LV Creation host, time hunk-virtual-machine, 2018-11-29 01:39:44 +0800
- LV Status available
- # open 0
- LV Size 400.00 MiB
- Current LE 100
- Mirrored volumes 2
- Segments 1
- Allocation inherit
- Read ahead sectors auto
- - currently set to 256
- Block device 252:8
- --- Segments ---
- Logical extents 0 to 99:
- Type raid1
- Monitoring monitored
- Raid Data LV 0
- Logical volume mirrorvol_rimage_0
- Logical extents 0 to 99
- Raid Data LV 1
- Logical volume mirrorvol_rimage_1
- Logical extents 0 to 99
- Raid Metadata LV 0 mirrorvol_rmeta_0
- Raid Metadata LV 1 mirrorvol_rmeta_1
测试Linaer/Striped LV
准备多个disk
在测试环境中添加了4个virtual disk,size都是10GB。
先用这4个virtual disk创建一个type为linear,size为20GB的Linaer LV,从后面查询LV的详情可以看出,这个LV实际上跨了3个PV。
- root@hunk-virtual-machine:/home/hunk# pvcreate /dev/sd[cdef]
- Physical volume "/dev/sdc" successfully created
- Physical volume "/dev/sdd" successfully created
- Physical volume "/dev/sde" successfully created
- Physical volume "/dev/sdf" successfully created
- root@hunk-virtual-machine:/home/hunk# vgcreate VolGroup1 /dev/sd[cdef]
- Volume group "VolGroup1" successfully created
- root@hunk-virtual-machine:/home/hunk# lvcreate -L 20G -n linnervol VolGroup1
- Logical volume "linnervol" created.
- root@hunk-virtual-machine:/home/hunk# mkfs.ext4 /dev/VolGroup1/linnervol
- root@hunk-virtual-machine:/home/hunk# mount /dev/VolGroup1/linnervol /volumetest
- root@hunk-virtual-machine:/home/hunk# df -h |grep linnervol
- /dev/mapper/VolGroup1-linnervol 20G 44M 19G 1% /volumetest
测试Linear LV
现在,用bonnie++来模拟IO,不停地向这个LV中写入数据。
- root@hunk-virtual-machine:/volumetest# bonnie++ -n 0 -u 0 -r `free -m | grep 'Mem:' | awk '{print $2}'` -s $(echo "scale=0;`free -m | grep 'Mem:' | awk '{print $2}'`*2" | bc -l) -f -b -d /volumetest/
- Using uid:0, gid:0.
- Writing intelligently...
在新窗口用bwm-ng来监控4个磁盘的IO速率,我们发现只有sdc上面有I/O请求,然而其他disk都很空闲,看着sdc一个家伙在忙。
bwm-ng -i disk -I sdc,sdd,sde,sdf
- bwm-ng v0.6 (probing every 0.500s), press 'h' for help
- input: disk IO type: rate
- \ iface Rx Tx Total
- ==============================================================================
- sdc: 0.00 KB/s 12263.47 KB/s 12263.47 KB/s
- sdd: 0.00 KB/s 0.00 KB/s 0.00 KB/s
- sde: 0.00 KB/s 0.00 KB/s 0.00 KB/s
- sdf: 0.00 KB/s 0.00 KB/s 0.00 KB/s
- ------------------------------------------------------------------------------
- total: 0.00 KB/s 12263.47 KB/s 12263.47 KB/s
我们继续查看LV中写入数据的量,直到写入数据超过10G时,发现sdc已经不再处理I/O请求了,因为数据已经塞满了嘛。而sdd开始继续处理持续的I/O请求。在写入数据10G多一点的时候,中间实际上有个过度过程,就是sdc和sdf都在处理I/O,这个是因为缓冲造成的。
- root@hunk-virtual-machine:/home/hunk# df -h |grep linner
- /dev/mapper/VolGroup1-linnervol 20G 11G 8.1G 57% /volumetest
- bwm-ng v0.6 (probing every 0.500s), press 'h' for help
- input: disk IO type: rate
- | iface Rx Tx Total
- ==============================================================================
- sdc: 0.00 KB/s 0.00 KB/s 0.00 KB/s
- sdd: 0.00 KB/s 12263.47 KB/s 12263.47 KB/s
- sde: 0.00 KB/s 0.00 KB/s 0.00 KB/s
- sdf: 0.00 KB/s 0.00 KB/s 0.00 KB/s
- ------------------------------------------------------------------------------
- total: 0.00 KB/s 12263.47 KB/s 12263.47 KB/s
测试Stripe LV
移除前面使用的Linear LV
- root@hunk-virtual-machine:/home# lvremove /dev/VolGroup1/linnervol
- Do you really want to remove and DISCARD active logical volume linnervol? [y/n]: y
- Logical volume "linnervol" successfully removed
创建一个条带化的LV
- root@hunk-virtual-machine:/home# lvcreate -L 20G --stripes 4 --stripesize 256 --name stripevol VolGroup1
- WARNING: ext4 signature detected on /dev/VolGroup1/stripevol at offset 1080. Wipe it? [y/n]: y
- Wiping ext4 signature on /dev/VolGroup1/stripevol.
- Logical volume "stripevol" created.
- root@hunk-virtual-machine:/home# lvdisplay /dev/VolGroup1/stripevol -m
- --- Logical volume ---
- LV Path /dev/VolGroup1/stripevol
- LV Name stripevol
- VG Name VolGroup1
- LV UUID z0MGOg-g6JL-hiE8-9Gt0-RZAJ-K29m-I6tcrS
- LV Write Access read/write
- LV Creation host, time hunk-virtual-machine, 2018-11-27 01:45:41 +0800
- LV Status available
- # open 0
- LV Size 20.00 GiB
- Current LE 5120
- Segments 1
- Allocation inherit
- Read ahead sectors auto
- - currently set to 4096
- Block device 252:6
- --- Segments ---
- Logical extents 0 to 5119: #Striped LV映射的PE均匀分布在了4个PV上
- Type striped
- Stripes 4
- Stripe size 256.00 KiB
- Stripe 0:
- Physical volume /dev/sdc
- Physical extents 0 to 1279
- Stripe 1:
- Physical volume /dev/sdd
- Physical extents 0 to 1279
- Stripe 2:
- Physical volume /dev/sde
- Physical extents 0 to 1279
- Stripe 3:
- Physical volume /dev/sdf
- Physical extents 0 to 1279
- root@hunk-virtual-machine:/home# mkfs.ext4 /dev/VolGroup1/stripevol
- mke2fs 1.42.13 (17-May-2015)
- Creating filesystem with 5242880 4k blocks and 1310720 inodes
- Filesystem UUID: 51dbdea0-48fc-4324-9974-42443e424aa0
- Superblock backups stored on blocks:
- 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
- 4096000
- Allocating group tables: done
- Writing inode tables: done
- Creating journal (32768 blocks): done
- Writing superblocks and filesystem accounting information: done
- root@hunk-virtual-machine:/home# mount /dev/VolGroup1/stripevol /volumetest/
- root@hunk-virtual-machine:/home# df -h |grep stripe
- /dev/mapper/VolGroup1-stripevol 20G 44M 19G 1% /volumetest
用同样的方法测试这个条带化的LV,不过这里的测试比较粗糙,不仅忽略了很多测试要素,前面对Linear LV的测试中bwm的I/O速率是每0.5s一次的采样值,而这里Striped LV的取的I/O速率是30s内的均值。不过我们这里并不是想得到准确的I/O速率,就先不考虑这些因素吧。明显能看出来4个disk在并行的处理I/O请求,也就是给Striped LV的I/O请求最终被分散到了多个底层disk上面,这样聚合的I/O效率必然会高出好几倍。
bwm-ng -i disk -I sdc,sdd,sde,sdf
- bwm-ng v0.6 (probing every 0.500s), press 'h' for help
- input: disk IO type: avg (30s) --取30S内采用的均值
- / iface Rx Tx Total
- ==============================================================================
- sdc: 0.13 KB/s 10010.92 KB/s 10011.05 KB/s
- sdd: 0.00 KB/s 10174.32 KB/s 10174.32 KB/s
- sde: 0.00 KB/s 6563.85 KB/s 6563.85 KB/s
- sdf: 0.00 KB/s 6113.09 KB/s 6113.09 KB/s
- ------------------------------------------------------------------------------
- total: 0.13 KB/s 32862.18 KB/s 32862.32 KB/s