1. This forum section is a read-only archive which contains old newsgroup posts. If you wish to post a query, please do so in one of our main forum sections (here). This way you will get a faster, better response from the members on Motherboard Point.

Other SAS disks in T5140

Discussion in 'Sun Hardware' started by gt, Aug 29, 2012.

  1. gt

    gt Guest

    Hello all,
    Maybe an obvious question, but does anyone know if we can use any old SAS disk in a Sun T5140? The Sun branded stuff is expensive, and can be had for half the price at Amazon in a HP bracket, will that work electronically (i.e the disk interface accepts it) and physically, i.e the bracket will slot into the server?

    Cheers

    Garry
     
    gt, Aug 29, 2012
    #1
    1. Advertisements

  2. gt

    Michael Guest

    Hi,

    That should work fine, the downside is if there is some firmware issues
    which is if its a Sun branded would be patched by Sun just like the HP
    is patched by HP.

    Ebay has lots of SUN branded drives aswell, both new and old. Samving
    money on bying a used drives is seldom saving unless its for private use.

    /michael
     
    Michael, Aug 29, 2012
    #2
    1. Advertisements

  3. I haven't seen any disk from Sun that had their own custom firmware on
    it (unlike NetApp, HP, Dell, etc), even if the disks were labelled as
    "Sun" disks. Those 3 above are the ones that typically have
    restrictions on using their disks and only their disks. (perhaps IBM
    and EMC as well).

    Doesn't mean that they didn't start doing it in the later hardware models.

    OOTH, the disk tray almost certainly won't fit, everybody has their
    own, and sometimes, that is more to obtain (from the rarer ones) than
    just sourcing the disk new in bracket from the maker.

    But now-a-days, you should be able to find just about anything used
    pretty cheaply.
     
    Doug McIntyre, Aug 29, 2012
    #3
  4. gt

    Nomen Nescio Guest

    Is the reverse a problem? Is it a problem to use HP, Dell, or IBM disks in a
    Sun server?
    Big SCSI drives are still really expensive unfortunately.
     
    Nomen Nescio, Aug 29, 2012
    #4
  5. gt

    Michael Guest

    Hi,
    NatApp did have really special firmware that I know for long time ago
    but I was not thinking of special firmware more "freezed" firmware.
    Sun may test some drives and then decide to use one of them, then they
    will request that all drives supplied to have the same firmware. Then
    the vendor will send information to Sun whenever there is a firmware
    issue and Sun will chose to update the drives or not using Sun patch
    policy. The Sun fireware could be a fork or just a selected version
    which is the most common way of doing it atleast what I know.

    Small companies(where I have been) usually install the selected firmware
    by themself and that could mean downgrade the firmware.
    Used drives is not something to use, especially used tape drives :)

    /michael
     
    Michael, Aug 29, 2012
    #5
  6. gt

    Michael Guest

    Hi,
    Today I don't think so, in the past NetApp atleast had special firmware
    that did not work well with atleast adaptec but this was in the 4GB era :)
    There are no "big", 300GB is the biggest I have seen or?

    But yes, they cost alot per MB.

    /michael
     
    Michael, Aug 29, 2012
    #6
  7. 541-2123 Silver Marlin Bracket

    John
     
    John D Groenveld, Aug 29, 2012
    #7
  8. I don't think I've used an OEM branded disk in a Sun machine
    specificly, but I don't remember any controller from Sun requiring
    certain disks. I've used branded OEM firmware disks elsewhere with
    zero detectable problems.
     
    Doug McIntyre, Aug 30, 2012
    #8
  9. gt

    Nomen Nescio Guest

    Thanks Doug. I need some big drives and it's good to know I don't have to
    confine my search to Sun branded drives (although in my experience they are
    excellent).
     
    Nomen Nescio, Aug 31, 2012
    #9
  10. gt

    gt Guest

    Thanks to all, will give the HP disks a go, a lot cheaper than our Sun quote.
     
    gt, Aug 31, 2012
    #10
  11. gt

    Anonymous Guest

    Hi,

    Sometimes I see those brands selling cheaper. Sun branded SCSI drives are
    always the most expensive for some reason. Are they any better? I only have
    Sun servers so I have no idea if those other drives are good. Sun enterprise
    drives seem to outlast the companies that bought them these days.
    That's what I meant. Big for SCSI.
    Yes, and they just cost alot per disk too.
     
    Anonymous, Aug 31, 2012
    #11
  12. gt

    ChrisQ Guest

    From experience here, you can use just about any scsi drive in a sun
    machine,
    you just have to run format to format / bad block and label the disk. For a
    146Gb drive, it does take several hours per drive, as the bad block scan is
    thorough. You can run several terminals worth of format at once, however.

    All drives have production tolerances and media error counts and it wouldn't
    surprise me to find that the sun badged drives are specially screened for
    quality and may even be burned in for a number of hours by the vendor...

    Regards,

    Chris
     
    ChrisQ, Aug 31, 2012
    #12
  13. Just for a big drive, you could try a regular SATA drive.
    SAS controllers can use those too.

    Dennis
     
    Dennis Grevenstein, Sep 1, 2012
    #13
  14. gt

    Michael Guest

    Hi,
    I don't think you can install a SATA in a T5140, there are different
    mating connectors in order to prevent wrong disks to be inserted.


    If you want extra disk, alot then I can recommend a external SAS chassis
    and a LSI HBA, I have just intalled two on our T2000

    Very nice build, only one powersupply so I have added two of these(two
    chassis and two BHA) and mirror in between then using ZFS!

    http://www.stardom.com.tw/sohotank_st8_feature.html
    http://www.lsi.com/products/storagecomponents/Pages/LSISAS9207-8e.aspx

    Cost about 2000 EURO in total then add SATA drives of your choice, may
    be two in each chassi same size and then place one drive in the mirror
    in each chassis and then one drive in each chassis as hot spare. Then
    you connect BHA-1 to chassis 1 and HBA-2 to chassis two.

    Or you could skip one BHA and only use 4 slots in each chassis.

    One SAS cable connects to four drives by BHA channel 0 -> disk1... HBA
    channel 3 -> disk 4 which means that if one drive locks up it will not
    lock up the other drives like SATA extenders will do.



    /michael
     
    Michael, Sep 1, 2012
    #14
  15. gt

    Nomen Nescio Guest

    Thanks alot for the info. I knew I had to format and lable but I did not
    know it was important to run the bad block check.
    I would guess they probably are screened, I've had excellent results with
    Sun branded drives but then I have nothing to compare it to except consumer
    PC drives and while I have had failures they're pretty good considering how
    much running time mine get.
     
    Nomen Nescio, Sep 2, 2012
    #15
  16. gt

    ChrisQ Guest


    Compared to windows, the format utility has a lot a of capability and
    has been
    around more or less unchanged since sun 3 days, which shows how far
    ahead of
    the pack they were at the time. I usually start by selecting format from
    the main
    menu, which formats and bad blocks the drive, then if any doubt about
    the drive,
    select analyse from the main menu, then compare, which writes the disk,
    reads back
    and compares results. That does destroy any data on the drive, but there
    are other
    options that don't. Once all that's done, you can go back to menu top
    level and
    use other menu options to partition the drive and write the label. I use
    a lot
    of s/user drives here and experience suggests that if you get any added
    bad blocks
    reported from the scan, the drive is on it's way out and probably should be
    binned. However, drives also get corrupted by system errors and it seems
    a waste
    to scrap them when a simple reformat would fix the problem...

    Regards,

    Chris
     
    ChrisQ, Sep 2, 2012
    #16
  17. actually, back in the Sun3 days one needed a disktab entry
    to format the drive, because format could not autodetect the
    disk geometry. At some point I got so annoyed that I attached
    the disk in question to a SPARC running Solaris to format
    it and then reattached the disk to the 3. At least this was a
    feature that it really lacked.

    Dennis
     
    Dennis Grevenstein, Sep 2, 2012
    #17
  18. gt

    ChrisQ Guest

    In those days, there was a list of drives, some of which are still
    there:
    Yes, it was a bit fiddly if you were trying to use a non sun drive, but
    all the capability was there to do it right down to selection of cylinders,
    tracks and heads. Never really understood why that was necessary with a scsi
    drive, which just presents a block count, but probably for historical
    reasons.

    Iirc, there were still references to and code for the 68k suns in quite
    late
    versions of solaris, none of which would run on 68k, but I just assumed
    that
    the engineers there had a finely tuned sense of the company history :)...

    Regards,

    Chris
     
    ChrisQ, Sep 2, 2012
    #18
  19. gt

    DoN. Nichols Guest

    While this is not a problem with *new* drives, I have found some
    examples of FC-AL drives which had been in some other maker's hardware
    RAID arrays, which use a non-standard sector size -- instead of the
    standard 512 byte sectors, they use (IIRC) 540 bytes per sector --
    presumably to store recovery data.

    Running format the normal way just does not work -- it fails
    every time. But -- if you run into one of these, there *is* a way to
    fix it. The following is assuming Solaris 10, FWIW.

    Start by running "devfsadm -C -c disk" to make sure that there
    is a device name for the drive (assuming that it is going in a SCSI ID
    which has not previously been used.) It also makes sure that there is
    not another WWN (FC-AL specific) pointing to the same slot.

    Then -- run format with a "-e" (expert) option.

    When you run format, it will spend a while spewing out quite a
    few repeats of something like this:

    ======================================================================
    Aug 9 22:47:20 Sysname FCP: WWN 0x2100000c5019659d reset successfully
    Aug 9 22:47:25 Sysname scsi: [ID 243001 kern.warning] WARNING: \
    /pci@8,600000/SUNW,qlc@4/fp@0,0 (fcp1):
    ======================================================================

    When it finally is ready to accept a command, issue the command
    "type", and enter '0' as the type. It will then come back with:

    ======================================================================
    Specify disk type (enter its number): 0
    Must reformat device to 512-byte blocksize. Continue? y
    Formatting...
    ======================================================================

    It does not tell you how long to expect to wait. With a 73 GB
    disk (I had to do this to twenty of these in one session) it took just a
    little less than an hour.
    Good Luck,
    DoN.
     
    DoN. Nichols, Sep 3, 2012
    #19
  20. gt

    DoN. Nichols Guest

    The reason for that is that early drives on Sun-2 and Sun-3 were
    either MFM or ESDI drives, connected to cards which made them appear as
    SCSI drives -- but the meta-information was not there on the disk as
    it is on true SCSI drives. These cards could handle two or up to a
    maximum of eight drives, which would be specified by the "d?" number in
    the Solaris /dev/dsk/c?t?d?s? names, and were also why the /dev/sd??
    numbers were sometimes so widely spaced.
    Interesting. SunOs 4.1.1 (the last available for the Sun3
    machines) had a problem with disks over 2GB in size. Most of the driver
    was right, but when it came time to use a substitute block to replace a
    bad block (the substitutes were located in the last couple of cylinders,
    normally) the code which specified the replacement sector used an older
    (narrower) set of commands in the SCSI command set, which truncated the
    block number to the width necessary to handle up to 2 GB. The result
    was that every write to a replaced block would instead overwrite
    something in the first 2 GB of disk -- usually something critical. :)

    Sun offered a patched version of the kernel for SunOs 4.1.1 for
    the SPARC systems -- but not for the Sun3 and Sun3x systems (68020 and
    68030 CPUs). This, I suspect, was to encourage people to migrate to
    SPARC systems -- which was fine for a serious company using the sytems,
    but for a hobbist using the Sun3 systems, it was bad news. :)

    Enjoy,
    DoN.
     
    DoN. Nichols, Sep 3, 2012
    #20
    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.