White Papers

crw------- 1 root root 10, 63 Jun 4 14:21 controlif

brw-rw---- 1 root disk 253, 3 Jun 10 15:31 mpio

brw------- 1 root root 253, 0 Jun 4 14:22 VolGroup00-LogVol00

brw------- 1 root root 253, 1 Jun 4 14:21 VolGroup00-LogVol01

17) Verify the configuration using the ‘dd’ command. Test that an I/O load is distributed

across the two Ethernet paths. Open the PS Group Manager and view the volume

named mpio. Open the ‘Connections’ tab and verify that both Ethernet IP addresses

are listed. Take note of the ‘MB Read’ column, which should be ‘0 MB’ for each

interface.

Execute this dd command:

# dd if=/dev/mapper/mpio of=/dev/null bs=4k count=1k

iflag=direct

1024+0 records in

1024+0 records out

4194304 bytes (4.2 MB) copied, 0.231391 seconds, 18.1 MB/s

This command moves 4MB through the iSCSI connection paths. Check the ‘MB

Read’ column again in the PS Group Manager. Each IP address should read ‘2 MB’,

indicating that the traffic was split across the two connection paths.

Debug tips:

 Start with an empty iscsiadm database = no records

 Logout of targets and delete records

 Start with an empty device-mapper database

 multipath –F (if it fails, use ‘dmsetup suspend’, ‘dmsetup remove’)

Task #6 - Mount an MPIO volume using the persistent name

In the previous section, multi-path I/O was configured on the SLES server. This

involved changes to the iscsid and multipathd configurations. A persistent name was

defined for the volume and the multipath connection was verified using the ‘dd’

command, which performs I/O operations on block devices. Now it’s time to mount

the ‘mpio’ volume as a filesystem.

Note: To ensure that iSCSI volumes being managed by device mapper are correctly

and consistently associated with a mount point at boot time, a persistent name should

be used. The procedure for defining a persistent name was covered in the previous

task

1) Create mount point:

# mkdir /mnt/mpio

2) Configure the volume to be mounted at boot time by adding the following line to

/etc/fstab:

/dev/mapper/mpio /mnt/mpio ext3 _netdev,defaults 0 0

3) Make a filesystem on the new volume. As stated before, the LVM tools could also

be used to initialize the volume and assign the space to a logical volume: