UNSOLVED

kruegerp

updated

16 years ago

K

kruegerp

1 Message

0

20699

March 2nd, 2011 06:00

MD3200i I/O errors

I'm using a MD3200i with two controllers in a CentOS 5.5. The iscsi-ports of the controllers are used via multipath

mpath1 (36842b2b000633046000003903b344cf8) dm-2 DELL,MD32xxi
[features=3 queue_if_no_path pg_init_retries 50][hwhandler=1 rdac][rw]
\_ round-robin 0 [prio=400][active]
\_ 23:0:0:0 sdg 8:96 [active][ready]
\_ 22:0:0:0 sdh 8:112 [active][ready]
\_ 26:0:0:0 sdi 8:128 [active][ready]
\_ 21:0:0:0 sdr 65:16 [active][ready]
\_ round-robin 0 [prio=0][enabled]
\_ 20:0:0:0 sdd 8:48 [active][ghost]
\_ 24:0:0:0 sdl 8:176 [active][ghost]
\_ 25:0:0:0 sdm 8:192 [active][ghost]
\_ 19:0:0:0 sdp 8:240 [active][ghost]

But I get continuos the following error messages

Mar 2 15:39:05 tensor kernel: Buffer I/O error on device sdc, logical block 0
Mar 2 15:39:05 tensor kernel: Buffer I/O error on device sdc, logical block 4394227967
...... also the other devices
Mar 2 15:39:06 tensor kernel: end_request: I/O error, dev sdg, sector 32
Mar 2 15:39:06 tensor kernel: end_request: I/O error, dev sdh, sector 32
Mar 2 15:39:07 tensor kernel: end_request: I/O error, dev sdi, sector 32
....... also the othe devices

The block and sector numbers are constant.

Can You please help me to solve the problem?.

Peter


  • mattslavin

    55 Posts

    480

    0

    Posted March 4th, 2011 06:00

    We get this alot, and its because, I feel the drivers or kernel paging is not the same as it is in Redhat 5.5 as it is for CentOS.

    I can tell you CentOS, isn't supported or validated to work with the array.
    http://support.dell.com/support/edocs/systems/md3200i/en/matrix/matrix.pdf

    Redhat 5.5 and SuSe will work perfectly with the device.

    Now in saying that we have been able to make the device work with other non-supported systems like Gentoo and such. First verify that linuxrcac driver is not being
    used. Run "rpm -q linuxrdac" and remove it if its there. You will use the DM Multipath and Open iSCSI for driver setup.

    Then follow the linux install manual located here http://support.dell.com/support/edocs/systems/md3200i/en/OM/PDF/MD3200i.pdf
    I can tell you that I have never seen Centos work correctly with the array, but I have seen BSD and debian systems work without a issue.
  • mrokkam1

    23 Posts

    480

    0

    Posted March 10th, 2011 19:00

    I believe that this is because the MD array is an active/active/passive array. i.e. A VD can be seen from both controllers but all IO will go through a single controller. So, if the server tries to send IO down the alternate path, you will receive errors like the above.

    Now, the errors should only be seen until the multipath driver loads and takes care of all path problems. So, if you are seeing it all the time, that may be due to some config problem.
  • mattslavin

    55 Posts

    480

    0

    Posted March 11th, 2011 04:00

    I agree there are many paths, and yes with the linuxrdac driver is normaly added to the kernel boot procedure. But the MD32xx uses the DM Multipath and it will not work in CentOS 5.5. You might try upgrading to the latest firmware released on 3/10/2011.

    When compairing Redhat 5.5 and CentOS builds I have seen two layouts to be the cause. The kernel is different and the mpt2sas modules are different. The answer is within how CentOS differs from those two code builds.