Supermicro P6DGU, flaky scsi disk IO - how to diagnose?

Discussion in 'Supermicro' started by Jonathan Rogers, Sep 30, 2004.

  1. About 6 months ago I started having SCSI disk problems on my
    Supermicro P6DGU. It would take the machine 2-3 minutes to recognize
    the disk at all on cold startup, then it would run fine. Then it took
    longer and longer, then I started getting disk errors. I figured the
    disk was bad and old, so I swapped it out for an identical spare. Same
    issues. Hmmm. I figured something else was wrong, so I completely
    traded disks as a test. Not even using the same technology (the old
    Maxtor Atlas 10K was a U160, the new Quantum can do U320 although of
    course the motherboard will only drive it at U160).

    But I'm still having problems, and in fact they're getting worse. More
    random errors, longer to start up from power-off, etc. I can't believe
    3 different drives are all bad, so I'm wondering if this 5-year-old SM
    motherboard is giving up the ghost as far as SCSI is concerned. But
    I'm at a loss to even begin to debug this...short of buying a
    replacement motherboard on Ebay and swapping the current one out.

    Ideas? Similar experiences? Thoughts on approaching this? I've like
    the P6DGU (this is destined for a server application, eventually) but
    at this point maybe I should just junk it and get another mobo?

    tia, -jonathan r-

    PS: The disk is attached directly to the SCSI connector on the
    motherboard, using the SM-supplied cable with an active terminator
    built in). It's the only device on the SCSI chain. I'm running a
    single PIII-1GHZ coppermine; the second CPU slot has a terminator in
    it. OS is Win98SE (yea, it's old, sue me). Not overclocked.

    PPS: It's not termination. The bus is terminated with an active
    terminator and was running fine before this with the same
    Jonathan Rogers, Sep 30, 2004
    1. Advertisements

  2. Try different cables. And check all power supply voltages.
    L David Matheny, Sep 30, 2004
    1. Advertisements

  3. Right, when there is no setup utility in the SCSI controller and jumpers on
    the drives to fiddle with, and your brain doesn't work, what else can one do?
    Can I have it please?
    I have always liked indestructable objects that have unlimited lifetime.
    Folkert Rienstra, Sep 30, 2004
  4. Jonathan Rogers

    Matt Reuther Guest

    You may want to check the board and make sure all the traces are still
    there. I actually had one of them burn. It caused odd problems like you
    described, but only once in a while when it the SCSI disks were under load.

    Matt Reuther, Oct 9, 2004
  5. It may also be the power supply.

    If the power supply is no longer putting out full current, then it would
    take longer and longer for the HD's to spin up to speed.
    Timothy Drouillard, Oct 9, 2004
  6. Following up on my original post and the suggestions provided.
    Motherboard has onboard monitors for all power supply voltages and
    they all turned out to be in spec. Ditto for for the SCSI cable, which
    I had tested and rebuilt (to reduce its length and to try a new active
    terminator). A go with a S.M.A.R.T diagnostic tool from Quantum
    confirmed the drive was happy and had no internal errors, but that the
    system occasionally couldn't even read/write to the drive's buffer. So
    I suspected something "upstream"....

    I pulled the motherboard out of the server completely and the
    problem's source instantly became clearer :). One of the UART ICs near
    the SCSI chips didn't look right; I tapped it lightly with a pen and
    it (literally) fell off the board into my lap. Ditto for the other two
    next to it. I suspect really poor soldering, which is unusual in my
    experience for Supermicro - normally of very high build quality. I'm
    sending the board out to get the capacitors replaced (some are bulging
    slighly - see for interesting perspective on this
    problem) and to get the UARTs resoldered. Unfortunately it's out of
    warranty, but this is the easiest route to get it fixed at this point.

    Many thanks to the helpful respondents (L.David, Timothy, Matt) on
    this thread for their suggestions on this - appreciated. NO thanks to
    the other respondent, whose life is apparently so empty that he feels
    a compulsion to fill its yawning void by spending all his waking hours
    on Usenet posting snide, unhelpful asides that serve only to show he's
    an ignorant Eurotrash wannabe who doesn't even fully read messages
    before launching into his infantile little
    I-am-so-smart-you-are-obviously-SO-stoopid routine. Or something like
    that. :)

    yours, /jr/

    Jonathan Rogers, Oct 10, 2004
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.