参考这一段：

Setting up Device Mapper Multipath

The purpose of Device Mapper Multipath is to enable multiple I/O paths to improve performance and provide consistent naming. Multipathing accomplishes this by combining your I/O paths into one device mapper path and properly load balancing the I/O. This section will provide the best practices on how to setup your device mapper multipathing within your Dell PowerEdge server.Verify that your device-mapper and multipath driver are at least the version shown below or higher:

rpm -qa | grep device-mapper-multipath

device-mapper-multipath-0.4.9-46.el6

Identify your local disks i.e. /dev/sda. Once your local disk is determined run the following command to obtain its scsi_id:

scsi_id --page=0x83 --whitelisted --device=/dev/sda

360026b900061855e000007a54ea53534

Open the /etc/multipath.conf file and locate and comment out the section below:

#blacklist {

# devnode “*”

#}

Once the scsi_id of your local disk has been retrieved, you must blacklist this scsi_id from being used as a multipath device within /etc/multipath.conf file and locate,uncomment,and modify the section below:

blacklist {

wwid

devnode "^(ram|raw|loop|fd|md|dm-|sr||scd|st)[0-9]*"

devnode "^hd[a-z]"

}

Uncomment your defaults section within your /etc/multipath.conf:

defaults {

udev_dir /dev

polling_interval 10

selector "round-robin 0"

path_grouping_policy multibus

getuid_callout "/sbin/scsi_id -g -u -s/block/%n"

prio_callout /bin/true

path_checker readsector0

rr_min_io 100

max_fds 8192

rr_weight priorities

failback immediate

no_path_retry fail

user_friendly_names yes

}

Locate the multipaths section within your /etc/multipath.conf file. In this section you will provide the scsi_id of each LUN/volume and provide an alias in order to keep a consistent naming convention across all of your nodes. An example is shown below:

multipaths {

multipath {

wwid

alias alias_of_volume1

}

multipath {

wwid

alias alias_of_volume2

}

}

Restart your multipath daemon service using:

service multipathd restart

Verify that your multipath volumes alias are displayed properly:

multipath -ll

Make sure iSCSI service starts upon boot using the command:

chkconfig multipathd on

Repeat steps 1-9 for all nodes.

[top]

Partitioning the Shared Disk

This section describes how to use Linux’s native partition utility fdisk to create  a single partition on a volume/virtual disk that spans the entire disk.

To use the fdisk utility to create a partition:

At the command prompt, type one of the following:

#> fdisk –cu /dev/

#> fdisk –cu /dev/mapper/

Where, is the name of the block device that you are

creating a partition on. For example, if the block device is /dev/sdb, type: fdisk –cu /dev/sdb

Device contains neither a valid DOS partition table, nor Sun, SGI or OSF disklabel

Building a new DOS disklabel with disk identifier 0x89058e03.

Changes will remain in memory only, until you decide to write them.

After that, of course, the previous content won't be recoverable.

Warning: invalid flag 0x0000 of partition table 4 will be corrected by w(rite)

The device presents a logical sector size that is smaller than

the physical sector size. Aligning to a physical sector (or optimal

I/O) size boundary is recommended, or performance may be impacted.

Command (m for help): n # To create a new partition

Command actione extendedp primary partition (1-4):

P # To create a primary partition

Partition number (1-4): 1

First sector (4096-xxxxxx, default 4096):

Repeat step 1 for all the disks.

Type the following to re-read the partition table and to be able to see the newly created partition(s)

#> partprobe

Or

#> service multipathd restart

Or

#> kpartx –a /dev/mapper/

Reboot the system if your newly created  partition is not

displayed properly.

[top]

Identifying ASM Disks to Setup Udev Rules **Applies to RHEL 6 and Oracle Linux 6 (if not using ASMlib)**

Red Hat Enterprise Linux 6/Oracle Linux 6 have the ability to use udev rules to ensure that the system properly manages permissions of device nodes. In this case, we are referring to properly setting permissions for our LUNs/volumes discovered by the OS. It is important to note that udev rules are executed in enumerated order. When creating udev rules for setting permissions, please include the prefix 20- and append .rules to the end of the filename. An example file name is 20-dell_oracle.rules

In order to set udev rules, one must capture the WWIDs of each disk to be used within your ASM device using the scsi_id command.

The command is as follows:

scsi_id --page=0x83 --whitelisted --device=/dev/sdX

where sdX is the name of your block device.

If one must run this command to capture multiple WWIDs, one could use the following for loop to do just that via the shell:

[root@rhel6 ~]# for i in sdb sdc sdd sde; do \ printf "%s %s\n" "$i" \ "$(scsi_id --page=0x83 --whitelisted --device=/dev/$i)"; done

Output:

sdb 360026b900061855e000008a54ea5356a

sdc 360026b9000618571000008b54ea5360b

sdd 360026b900061855e000008a54ea5356a

sde 360026b9000618571000008b54ea5360b

Once the WWIDs have been captured, create a file within the /etc/udev/rules.d/ directory and name it 20-dell_oracle.rules. A separate KERNEL entry must exist for each storage device and will require adding the WWID to the "RESULT==" field.

An example of what needs to be placed in the /etc/udev/rules.d/20-dell_oracle.rules file

#------------------------ start udev rule contents ------------------#

KERNEL=="dm-*", PROGRAM="scsi_id --page=0x83 --whitelisted --device=/dev/%k",RESULT=="360026b9000618571000008b54ea5360b", OWNER:="grid", GROUP:="asmadmin"

KERNEL=="dm-*", PROGRAM="scsi_id --page=0x83 --whitelisted --device=/dev/%k",RESULT=="360026b900061855e000008a54ea5356a", OWNER:="grid", GROUP:="asmadmin"

#-------------------------- end udev rule contents ------------------#

As you can see from the above, the KERNEL command looks at all dm devices, executes the PROGRAM which captures the WWID, if the RESULT of the WWID is matched, appropriately assign the grid user as the OWNER and the asmadmin group as the GROUP.