Adaptec 29320 Raid Setup - Need Help - SCO

This is a discussion on Adaptec 29320 Raid Setup - Need Help - SCO ; First of all, to anybody that purchases the Dell SC420 Power Edge servers to run SCO, be prepared for a fight. These servers SUCK!! The Adaptec 39320 controller that comes with the Dell doesn't work. It is a "special" Dell ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: Adaptec 29320 Raid Setup - Need Help

  1. Adaptec 29320 Raid Setup - Need Help

    First of all, to anybody that purchases the Dell SC420 Power Edge
    servers to run SCO, be prepared for a fight. These servers SUCK!! The
    Adaptec 39320 controller that comes with the Dell doesn't work. It is a
    "special" Dell card (**** THEM!!)that will not work with SCO. Also, I
    had to dig for the NIC drivers because Dell doesn't support those
    either on SCO at all.

    So...I bought an Adaptec 29320 retail and was able to get it to load
    SCO 5.0.6 as long as I had HostRaid turned off. If I turn HostRaid on
    for Raid 1 it gives me the "no root disk controller error" I want Raid
    1 that's why I bought the controller.

    Does anybody know what I need to do to make this P.O.S. work? I'm about
    ready to throw this thing through a window.

    Thanks in advance


  2. Re: Adaptec 29320 Raid Setup - Need Help



    "Steve M. Fabac, Jr." wrote:
    >
    > unixfox wrote:
    > >
    > > Your right Tony, I should have looked more carefully at the Adaptec
    > > support page. It does say *without* HostRaid enabled.
    > >
    > > So, I'm back to square one. What controller *DOES* work with Raid 1
    > > Ultra 320 drives? Is this even an option or do I need to build my own
    > > machines from now on?

    >
    > I have used Adaptec 2110 RAID controllers. They are only U160 but you will
    > not likely see the difference in performance unless you have very disk
    > intensive software.
    >
    > I have been building servers with SuperMicro system boards and have been
    > happy with the ones supporting Adaptec 2010 Zero Channel RAID adapters.
    > The 2010 converts the on-board Adaptec SCSI controller into RAID controllers.
    >
    > The last server contained a SuperMicro X5DL8-GG system board Xeon 2.8GHz 533MHZ FSB
    > (The 533MHz FSB Xeons are no longer available so I will be using different SuperMicro
    > system boards in the future.)
    >
    > That gives you U320 RAID using the updated (ne bootload) dpti driver:
    >
    > # pkginfo
    > preSVR4 dpti5 SCO UNIX 3.2.5.X Adaptec I2O SCSI Host Adapter Driver
    >


    OOPS,

    Forgot to say that there are some downsides to the Adaptec 2010s: If you enable the
    system monitoring daemons: /usr/dpt/dpteng, /usr/dpt/dptagent, ,/usr/dpt/dptelog and
    /usr/dpt/dptscom on a single CPU system, sar shows the system as no more then 90% idle
    when no users are logged on. (On a SMP system this is not a problem and idle is up to
    100% with no users on.) Additionally, You can't use a Windows dptmanager client to
    connect to the UNIX system to administer the system remotely. (Adaptec tech notes
    refers to this problem but offers no solution.)

    So I disable all the dpti daemons except /usr/dpt/dpteng and that allows failed and
    replaced RAID disk to be automatically detected and rebuilt.

    Lastly, I miss the Mylex DAC960 utilities that displayed a warning message on the system
    console when the RAID was degraded. To replace that, I created the following script to
    provide the same utility with the Adaptec 2010 and 2110S controllers. It works somewhat but
    it does not provide a blinking "ATTENTION!!" banner. The Mylex utility displayed a red
    background warning banner with blinking text to alert anyone viewing the console screen
    that the RAID is critical and has failed components. My scrip is clumsy with screen color
    and achieves a red background but I can't get the blinking text and it makes no attempt to
    preserver current screen colors and restore them upon exit.

    If anyone can suggest modifications to the script to get the screen color correct please
    respond with your suggestions.

    TIA

    # cat /usr/bin/checkraid
    :
    #
    # @(#) checkraid 1.1 10/21/02 smf
    #
    #Address Type Manufacturer/Model Capacity Status
    #---------------------------------------------------------------------------
    #d0b0t0d0 RAID 1 (Mirrored) ADAPTEC RAID-1 17501MB Optimal
    # d0b0t0d0 Disk Drive (DASD) IBM IC35L018UCD210-17501MB Optimal
    # d0b0t2d0 Disk Drive (DASD) IBM IC35L018UCD210-17501MB Optimal

    TERM=scoansi;export TERM
    /usr/dpt/raidutil d0 -L raid | awk -v pid=$$ ' BEGIN {
    ID[ "d0b0t0d0" ] = "(0,0,0,0)"
    ID[ "d0b0t1d0" ] = "(0,0,1,0)"
    ID[ "d0b0t2d0" ] = "(0,0,2,0)"
    ID[ "d0b0t3d0" ] = "(0,0,3,0)"
    ID[ "d0b0t4d0" ] = "(0,0,4,0)"
    ID[ "d0b0t5d0" ] = "(0,0,5,0)"
    system ( "tput -T scoansi blink > /tmp/raid.head" )
    system ( "echo \r ! ATTENTION !! ATTENTION !! ATTENTION !! ATTENTION !! ATTENTION !! ATTENTION ! >> /tmp/raid.head" )
    system ( "tput -T scoansi rmso >> /tmp/raid.head" )
    }

    $1 ~ /d0b0/ {
    if ( $2 == "RAID" ) {
    if ( $NF !~ "Optimal" ) {
    printf "RAID %s: %s %s\r\n", ID[ $1 ], $(NF - 1), $(NF)
    print pid > "/tmp/raid.pid"
    }
    else printf "RAID %s: %s\r\n", ID[ $1 ], $(NF)
    }
    if ( $2 == "Disk" ) {
    if ( $NF !~ "Optimal" ) {
    printf "Disk %s: %s %s\r\n", ID[ $1 ], $(NF - 1), $(NF)
    }
    else printf "Disk %s: %s\r\n", ID[ $1 ], $(NF)
    }
    }
    '>> /tmp/raid.$$
    awk -v "tmp=$$" ' /RAID/ { if ( $3 !~ "Optimal" ) {
    warn = "[=0E[=14F[=4G" << Note: the hollow box left is an esc so "esc[0m" etc. >>
    clearit = "[=0E[=7F[=0G" << I don't know it it wll translate properly via cut and paste>>
    printf "%s\n", warn > "/dev/console"
    system ( "cat /tmp/raid.head > /dev/console")
    system ( "echo \r > /dev/console")
    system ( "date > /dev/console")
    system ( "echo \r > /dev/console")
    system ( "cat /tmp/raid." tmp " > /dev/console")
    system ( "dtox /tmp/raid." tmp " > /tmp/raid.head")
    system ( "logger -f /tmp/raid.head -t RAID " )
    printf "%s\n", clearit > "/dev/console"
    }
    } ' /tmp/raid.$$


    check=`egrep -q "Degraded|Reconstruct" /tmp/raid.$$; echo $?`
    [ "$check" = "0" ] && {
    clearit="[=0E[=7F[=0G"
    tput -T scoansi blink > /dev/console
    echo $clearit > /dev/console
    tput -T scoansi rmso > /dev/console
    } || {
    if [ -f /tmp/raid.pid ]
    then
    awk -v "tmp=$$" ' /RAID/ { if ( $0 ~ "Optimal" ) {
    warn = "[=0E[=14F[=4G"
    clearit = "[=0E[=7F[=0G"
    printf "%s\n", warn > "/dev/console"
    system ( "cat /tmp/raid.head > /dev/console")
    system ( "echo \r > /dev/console")
    system ( "date > /dev/console")
    system ( "echo \r > /dev/console")
    system ( "cat /tmp/raid." tmp " > /dev/console")
    system ( "dtox /tmp/raid." tmp " > /tmp/raid.head")
    system ( "logger -f /tmp/raid.head -t RAID " )
    printf "%s\n", clearit > "/dev/console"
    }
    } ' /tmp/raid.$$
    rm /tmp/raid.pid

    fi
    }

    rm -f /tmp/raid.$$ /tmp/raid.head


    --

    Steve Fabac
    S.M. Fabac & Associates
    816/765-1670

  3. Re: Adaptec 29320 Raid Setup - Need Help

    (Steve M. Fabac, Jr.) 23.07.05 in /comp/unix/sco/misc:



    >I have used Adaptec 2110 RAID controllers. They are only U160 but you
    >will not likely see the difference in performance unless you have very
    >disk intensive software.


    >I have been building servers with SuperMicro system boards and have
    >been happy with the ones supporting Adaptec 2010 Zero Channel RAID
    >adapters. The 2010 converts the on-board Adaptec SCSI controller into
    >RAID controllers.


    >The last server contained a SuperMicro X5DL8-GG system board Xeon
    >2.8GHz 533MHZ FSB (The 533MHz FSB Xeons are no longer available so I
    >will be using different SuperMicro system boards in the future.)


    >That gives you U320 RAID using the updated (ne bootload) dpti driver:



    >Address Manufacturer/Model Write Cache Mode
    >----------------------------------------------------------------------
    >----- d0b0t4d0 SEAGATE ST336807LC Write Back
    >d0b0t0d0 ADAPTEC RAID-1 Write Back
    > d0b0t0d0 SEAGATE ST336807LC Write Back
    > d0b0t2d0 SEAGATE ST336807LC Write Back


    Isn't very risky to run "write back" on disks too?
    The RAID controler can get a battery backup unit, but
    if the disk looses the power, the journal may become broken,
    as it is assumed to be realy written.

    Is the performance gain by allowing unbuffered writ back on the disks
    worth that risc to get a corupt filesystem?
    Have tested it?

    Rainer

  4. Re: Adaptec 29320 Raid Setup - Need Help


    ----- Original Message -----
    From: "unixfox"
    Newsgroups: comp.unix.sco.misc
    To:
    Sent: Saturday, July 23, 2005 9:43 AM
    Subject: Re: Adaptec 29320 Raid Setup - Need Help


    > Your right Tony, I should have looked more carefully at the Adaptec
    > support page. It does say *without* HostRaid enabled.
    >
    > So, I'm back to square one. What controller *DOES* work with Raid 1
    > Ultra 320 drives? Is this even an option or do I need to build my own
    > machines from now on?
    >
    > I guess I've learned my lesson.


    Adaptec u320 controllers work fine, including full raid 0,1, 5, 10, etc...
    You just have to get a "real" controller, not the crap^h^h^h^hhost-based
    raid models.
    2130 for instance. I don't know if they have fixed the problem of booting
    from this family of controller with SMP though. By now they really should
    have.

    I've been using LSI controllers. (Megaraid 320-1, 320-2 & 320-2X myself
    although the entire line looks to be supported)
    The readme for the latest driver even lists some Dell models.
    ftp://ftp.sco.com/pub/openserver5/50....25_readme.txt

    Update the firmware and make sure you use the latest driver. I had problems
    with stock firmware a few months ago and cards in the distribution channel
    might still have an old firmware.
    The problems were basically that I could crash the drive by doing things
    that should be legal (removing drives, stressing the system while drives are
    missing or rebuilding, removing a drive while it's rebuilding, etc...).
    Meaning the whole point of having raid was lost.

    With the latest firmware and driver, I can't get it to fail anymore, but I
    do still have a half-problem. If I invalidate a drive, I can't get the
    controller to recognize the drive as a good useable drive when I plug it
    back in. Untill I did this to more drives and refused to accept that this
    many brand new drives were really bad just from unplugging them I had
    actually returned a few drives under warranty.

    Eventually I figured out I could use a seperate test box with the same
    family of controller and boot into the web bios to sort of reset the drive.
    I don't remember exactly what I had to do, possibly low-level format, which
    is generally not a smart thing to do. Maybe nothing more than configuring an
    array and then deleting it again. After that I plug the drive back into the
    same array that rejected it before, it sees it immediately as a new working
    drive and rebuilds it. This would be a real prolem if I only had one
    controller in one production box, but since I have a test box with a similar
    controller that's why I call it a half a problem. It's also not likely to be
    a problem in most real situations because generally you plug a drive in and
    it doesn't become invalid except when it really breaks, and you don't plug
    that same drive back in but get a new one. I've never had the card reject a
    new drive, only ones that were in an array and then marked bad.

    Pay attention to the install notes (link above) for the driver. They use a
    new numbering scheme that uses virtual bus (aka channel) numbers to allow up
    to 40 drives in an array. So unlike most other controllers the first logical
    drive is not 0,0,0,0 (adapter,bus,id,lun)

    Brian K. White -- brian@aljex.com -- http://www.aljex.com/bkw/
    +++++[>+++[>+++++>+++++++<<-]<-]>>+.>.+++++.+++++++.-.[>+<---]>++.
    filePro BBx Linux SCO FreeBSD #callahans Satriani Filk!


  5. Re: Adaptec 29320 Raid Setup - Need Help

    On Sat, 23 Jul 2005 22:34:00 +0100, Rainer Zocholl wrote
    (in message <9aQTLfnMgjB@zocki.toppoint.de>):

    > Isn't very risky to run "write back" on disks too?
    > The RAID controler can get a battery backup unit, but
    > if the disk looses the power, the journal may become broken,
    > as it is assumed to be realy written.


    It depends.

    IF you have adequate battery (or other) backup for the controller cache, and
    teh controller is properly built (including it's software), then it's fairly
    safe. In the event of power loss, the controller will retain data for pending
    writes and when power comes back up will write it to disk.

    Depending on your application, the performance gain may be anything from
    marginal (in which case you would be right to suggest write back is more
    danger than it's worth) to 'quite significant'. It's just one of the those
    tradeoffs you have to make when designing a system.


    But back to teh earlier question about working raid cards, at work we used to
    have a Mylex DAC960, but our current system is an IBM box and has one of
    their raid controllers. Both work fine with SCO, though the IBM one had a
    known bug when it shipped which was fixed by a firmware upgrade.

    In terms of U160 vs U320, in most applications I suspect that you'll run into
    host bus bandwidth limitations before you run out of SCSI bandwidth -
    especially if you have multiple SCSI busses.

    Simon


+ Reply to Thread