1

Hope i am not wasting anyone´s time, so apologies to all should that be the case with this question.

it is the first time i am playing with multipath, manged to setup multipath on a RHL5, the setup comprises of the following:

RHL5 Server has 2 x HBA - HBA1 Port1 connects to Fiber Switch 1 & HBA2 port1 connects to fiber switch 2

NETAPP FAS2040 Storage with one HBA with 2 ports, port 1 connects to Fiber Switch 1 & port 2 connects to Fiber Switch 2.

multipath -ll command show the following:

[root@nonameserver /]# multipath -ll
mpath0 (360a98000323766586b24422f6735626c) dm-2 NETAPP,LUN
[size=149G][features=1 queue_if_no_path][hwhandler=0]
\_ round-robin 0 [prio=2][active]
 \_ 4:0:0:0 sda 8:0   [active][ready]
 \_ 2:0:0:0 sdb 8:16  [active][ready]

My question is why is there two devices /dev/sda & /dev/sdb? on the storage there is only one LUN setup for the host.

Thank you all.

John
  • 9,070
  • 1
  • 29
  • 34
Vlad
  • 45
  • 6
  • Hi, browsing the site i found a similar question in here http://serverfault.com/questions/532644/mounting-a-disk-from-san-rhel-5-9 but i would like to have more opinions. thanks. – Vlad Oct 11 '13 at 14:54

2 Answers2

4

The /dev/sda device is the target LUN as seen through one HBA, the /dev/sdb device is the same LUN seen through the other HBA. That's how it works - different controllers means different devices. As the other answer you commented about notes, do not use sda or sdb - use the multipath device (/dev/mpath0).

John
  • 9,070
  • 1
  • 29
  • 34
  • Hi John, does it mean then that by using the /dev/mpath0 device it would fail over to either /dev/sda or dev/sdb depending on its availability? – Vlad Oct 11 '13 at 15:39
  • In simple terms and at a high level, yes, that is what multipathing does. – John Oct 11 '13 at 16:23
1

I hope I understand your question. This is how I do it:

$ iscsiadm --mode discovery --type sendtargets --portal <IP:PORT>

$ iscsiadm -m node -l (login to all)

multipath -ll -v2 (This will show all devices currently connected)

ls -la /dev/s* (e.g. Look for the iscsi device with latest date and time)

scsi_id -g -u -s /block/sxxxx (e.g. Replace sxxx with the block name)

vim /etc/multipath.conf

multipath {

             wwid      <put the output of scsi_cmd here>
             alias     <your-alias>
             path_selector "round-robin 0"
             path_grouping_policy multibus
             rr_weight  uniform
             no_path_retry 5
             rr_min_io   100
}

service multipath restart

multipath -ll -v2 (alias listed now)

udevcontrol reload_rules

(Now edit your fstab and add /dev/mapper/<alias>p1 /mntpoint ext3 _netdev 0 0

mkdir /mntpoint

parted /dev/mapper/<alias>

parted> mklabel gpt

parted> print

parted> name 1 alias

parted> print

parted> quit

ls /dev/mapper

mkfs.ext3 -m 0 -F /dev/mapper/<alias>p1

tune2fs -c 0 -i 0 /dev/mapper/<alias>p1

e2label /dev/mapper/<alias>p1 mntdir

mount -a

Optional: Add to NFS Server

vim /etc/exports

(add mnt dir to exports if you want to share it on NFS)

Engineer2021
  • 601
  • 8
  • 25
  • why make partition on dm multipaths? a lun is storage logical partition and you did other partion on the server side, i think do that is waste your time – c4f4t0r Jun 07 '14 at 10:22
  • @c4f4t0r: this is how it is recommended in the multipath manual – Engineer2021 Jun 07 '14 at 13:31
  • @If you have a redhat network account and you ask to the author of the manual, nobody can't tell you why partitioning a lun is required – c4f4t0r Jun 07 '14 at 13:40