multipath iscsi on debian linux with an md3000i

Bryan Mclellan bryanm at widemile.com
Wed May 28 18:39:18 CDT 2008


So I have an MD3000i with two controllers, each with two network interfaces plugged into a switch connected to a debian etch box (it's a bit more complicated than that, but I believe those parts are irrelevant). Right now the client only has one network interface on the iscsi switch but has an alias on each of the iscsi subnets.

After resolving my open-iscsi kernel/version problems [1] I've been able to access the volume okay but I can only access it via two nodes at a time. It appears that these two nodes are the two on the preferred controller, as when I switch the preferred controller via MDSM, I can only access the volume across the other two nodes.

        # iscsiadm -m session
        tcp: [3] 10.0.8.11:3260,2 iqn.1984-05.com.dell:powervault.6001c23000d59fc6000000004754447a
        tcp: [4] 10.0.9.10:3260,1 iqn.1984-05.com.dell:powervault.6001c23000d59fc6000000004754447a
        tcp: [5] 10.0.8.10:3260,1 iqn.1984-05.com.dell:powervault.6001c23000d59fc6000000004754447a
        tcp: [6] 10.0.9.11:3260,2 iqn.1984-05.com.dell:powervault.6001c23000d59fc6000000004754447a

        # cat /proc/partitions
        (some snipping)
        major minor  #blocks  name
           8    16  584888320 sdb
           8    17  584886456 sdb1
           8    32  584888320 sdc
           8    33  584886456 sdc1
           8    48  584888320 sdd
           8    64  584888320 sde

Notice how there's only a partition on two of the disks. Running fdisk against two nodes operates fine, but against the other two it fails.

I also get a bunch of I/O errors on the other disks

        end_request: I/O error, dev sdd, sector 0
        end_request: I/O error, dev sde, sector 0
        end_request: I/O error, dev sdd, sector 0
        printk: 7 messages suppressed.
        Buffer I/O error on device sdd, logical block 0
        Buffer I/O error on device sdd, logical block 1
        Buffer I/O error on device sdd, logical block 2
        Buffer I/O error on device sdd, logical block 3
        end_request: I/O error, dev sdd, sector 0
        Buffer I/O error on device sdd, logical block 0
        end_request: I/O error, dev sde, sector 0
        Buffer I/O error on device sde, logical block 0
        Buffer I/O error on device sde, logical block 1
        Buffer I/O error on device sde, logical block 2
        Buffer I/O error on device sde, logical block 3
        end_request: I/O error, dev sde, sector 0

Is this expected? If I set up multipath correctly and pull the preferred controller should the other two block devices magically start working? That is, is having two controllers an active/passive configuration? Can you balance the load across two controllers, or four nodes in total or is the multipath only for redundancy?

I've had some problems with controller 1, originally it came with a bad battery and dell replaced it, now it needs to have its time synced daily (Case #4235390 and #4776180 if any dell folks are interested). Perhaps that has something to do with this? Having only one of these I have nothing to compare to so I'm hoping someone else has some linux multipath experience.

Bryan McLellan
Systems Administrator
Widemile: The Next Wave of Online Optimization
bryanm at widemile.com
206.985.7171 ext. 117 - office
206.607.7108 - cell





More information about the Linux-PowerEdge mailing list