Results 1 to 10 of 10
  1. #1
    Joined
    Jul 2005
    Posts
    19

    Down to 2 Drives, need advice

    Hello, I recently put together a new machine. MB is a GIGABYTE GA-880GA-UD3H AM3 AMD 880G SATA 6Gb/s.

    I purchased a WD6401AALS which was DOA. Comparing prices for a replacement I can get another WD6401AALS or WD6402AAEX for the same price from two respected vendors since one charges me tax and one doesn't. Since my MB supports SATA 6.0 should I try the AAEX which I've read mixed reviews on get another AALS which the 1st one was DOA?

    Any input would be appreciated.

  2. #2
    sttubs is offline Always learning something
    Joined
    Feb 2004
    Location
    Rock Falls, IL
    Age
    53
    Posts
    988

    Re: Down to 2 Drives, need advice

    If you're planning on a RAID array, the cache sizes should be the same. The WD6401AALS is still a very good harddrive despite your unfortunate DOA. You probably wouldn't notice any difference with 32mb vs 64mb cache. You could take advantage of your situation & go for a 1tb drive, their prices have come down nicely.

  3. #3
    Joined
    Oct 2009
    Posts
    53

    Re: Down to 2 Drives, need advice

    Western Digital does not recommend their Caviar Blacks for RAID arrays,
    because they do not support TLER (time-limited error recovery).

    WD's RAID Edition ("RE") HDDs support TLER:

    http://www.wdc.com/en/products/produ...sp?language=en


    If a controller polls a hard drive but that drive's firmware is
    busy doing error checking, the drive may not respond to the poll
    and the controller may drop that drive from a RAID array
    because the controller thinks the drive is dead.


    MRFS

  4. #4
    Joined
    Jul 2005
    Posts
    19

    Re: Down to 2 Drives, need advice

    Thanks for the heads up but I'm not planning on doing RAID.

    Right now I'm so frustrated I could scream. I ordered the 640 AAEX mentioned above as a replacement. It worked for 3 days then all the sudden started making a constant loud screaching sound, a few seconds later it crashed, totally dead now. Just long enough for me to get everything transfered over to it and configured the way I wanted.

    Anyway, I guess it's time for round number 3, I'm fed up with WD. I know it's a fluke but still annoying. I've bought many of them over the years, never had a bad one, now 2 in a row. Any ideas for the next one around the same budget, something dependable that I don't hear in the next room? Should I stick with WD or look elsewhere?

  5. #5
    sttubs is offline Always learning something
    Joined
    Feb 2004
    Location
    Rock Falls, IL
    Age
    53
    Posts
    988

    Re: Down to 2 Drives, need advice

    Personally I'd stick with WD. I've had too many good hdd's & only a handful of bad ones.

  6. #6
    Joined
    Jul 2005
    Posts
    19

    Re: Down to 2 Drives, need advice

    Yeah I'm going to try them again, was just thinking out loud during frustration. I've never had a bad one until the last 2 in a row. Monday they are over nighting me a replacement on their dime for my troubles. It'll be here Tuesday so hopefully number 3 is the charm. Sucks having 2 RMA's waiting to be credited at the same time though. This HD quest is making my budget build.....not so budget.

  7. #7
    Joined
    Oct 2009
    Posts
    53

    Re: Down to 2 Drives, need advice

    Quite often, it's poor handling by the shipper
    that is the main factor causing these failures.

    I once knew an experienced UPS employee
    who joked one day that "all UPS packages go by air"
    and he threw a small package across the room
    into the outgoing bin, as if it were a basketball.

    There were so many complaints about Newegg's
    packaging of OEM HDDs, that I ended up writing
    to Western Digital to tell them about these complaints.

    After that, Newegg's packaging noticeably improved.

    So, before you blame WD, consider the possibility that
    several other factors may explain the DOAs.


    MRFS

  8. #8
    Joined
    Jan 2009
    Posts
    65

    Re: Down to 2 Drives, need advice

    Quote Originally Posted by MRFS View Post
    Western Digital does not recommend their Caviar Blacks for RAID arrays,
    because they do not support TLER (time-limited error recovery).
    RE's are meant for enterprise. *Good* RAID cards are not so sensitive to drives doing background error checking. WD's that are actively background-error-checking will almost instantly respond to OS (or RAID) requests and then resume the checking once the host goes idle again.

    I've personally ran the following without a single dropout:
    8-drive RAID-5 of 1TB Caviar Greens for 3+ months.
    10-drive RAID-6 of 1TB Caviar Greens for 3+ months.
    6-drive RAID-5 of 2TB Caviar Greens for 1+ month.

    The RAID card used for the above was an Areca ARC-1230 (still in use).

    A long time ago I ran a 6-drive RAID-5 of 250GB WD's. That array did fail, but it was due to legitimate dropouts / timeouts due to bad sectors / failed heads on two drives of the array. Those drives ran very hot and had a bad track record. I recovered the data by imaging the drives (around the errors) and manually destriping the images in software, correcting the missing parts with the parity of each stripe. Data recovery would have been simpler if not for the Promise RAID card having such crappy error handling / notification. The first drive had been failed for a solid week before I was made aware of it, and the subsequent failure occurred when I initiated the manual rebuild.

    TLER is not as necessary as everyone thinks. All it does is prompt the drive to give up on ECC and report a failed sector sooner than the RAID card's threshold for considering the drive unresponsive (dropped out). My current array has been up for a year (8x RE4-GP 2TB in RAID-5) and reports 0 'Media Error Count' for all drives. My move away from Caviar Greens was for the performance gain, not the TLER.

    Final data point for consideration: Drobo (a RAID in an external enclosure) sells their units with Caviar Greens. They are not even available with RE drives:
    http://www.drobostore.com/store/drobo/DisplayHomePage

    All of that said, even with my current use of RE4's and RAID, I *still* regularly back that array up to another set of disks. Regardless of what you're using, never consider a RAID with the only copy of your data as a backup regime. Save your money and buy Caviar Green drives, and use that saved money on your backup solution. That way you have your backup, and if by some fluke / chance you get a drive drop out and kill the array, just rebuild the thing from scratch and restore from your backup.
    Allyn Malventano, CTNC, USN
    PCPer Storage Editor

  9. #9
    Joined
    Oct 2009
    Posts
    53

    Re: Down to 2 Drives, need advice

    > *Good* RAID cards are not so sensitive to drives doing background error checking.


    Good post, Allyn: however, what's a "good" RAID card
    and what's a "not so good" RAID card, and how is one to know?

    You saved money on Caviar Greens, but then you paid a premium
    for an Areca controller (one of THE most expensive).


    From what I have been gleaning from the Comments at Newegg,
    the dropped member problem appears to be most prevalent
    when the on-board RAID functions of chipsets are used
    INSTEAD OF expensive add-on controllers.

    You are also recommending that users backup their backups --
    adding more cost to their hardware budgets.


    For myself, I approach this issue in an analytical fashion:
    there are X many HDDs that can be wired to Y many SATA & SAS ports:
    the arithmetic product of X times Y is a large number, and
    certain combinations are known to fail.


    If Drobo are using Caviar Greens, I would expect that their
    engineers have determined that their controller logic is not
    prone to dropping member HDDs, or simply programmed it that way.


    As for our experiments, I can definitely say that enabling polling
    on a Highpoint RocketRAID 2322 definitely causes a RAID 0 array
    to fail, quite often. As soon as I disabled that polling function,
    there have been no array failures -- NONE!

    And, this difference occurred with Western Digital RAID Edition HDDs!!

    So, even with WD's TLER, a "bad" controller (using your terminology)
    can decide to drop a member disk and do so erroneously.


    Perhaps what the IT industry needs is a standard for configuring
    "polling" functionality, and what to do if/when polling requests
    are not answered.

    If I were programming that logic, I would make sure that
    User / System Administrator has some reasonable choices,
    e.g. when to drop a member drive, when NOT to drop a member drive,
    and so on.

    And, I don't think it's too much to ask that such polling be
    enabled in such a way that it occurs only when a given
    HDD is not already busy -OR- be configurable by allowing
    the User / System Administrator to disable polling entirely,
    as I have done with the RR2322.


    I hope this helps.


    MRFS

  10. #10
    Joined
    Jan 2009
    Posts
    65

    Re: Down to 2 Drives, need advice

    > Good post, Allyn: however, what's a "good" RAID card
    and what's a "not so good" RAID card, and how is one to know?

    I hate to say it but it's Areca or bust until I see something equally good come along. I've been burned in one way shape or form by nearly every other maker. Highpoint is lacking in their driver / firmware implementation, which I bet was causing those dropouts with polling.

    > You saved money on Caviar Greens, but then you paid a premium
    for an Areca controller (one of THE most expensive).

    At the time I got a heck of a deal on a 12-port Areca. There was a distributor selling a batch of used ones for $400 for a 6-month period through pricegrabber.

    > From what I have been gleaning from the Comments at Newegg,
    the dropped member problem appears to be most prevalent
    when the on-board RAID functions of chipsets are used
    INSTEAD OF expensive add-on controllers.

    Yeah, chipsets are picky. My ICH10R re-checks my OS SSD RAID if the system crashes in any sort of funny manner.

    > You are also recommending that users backup their backups --
    adding more cost to their hardware budgets.

    No, I'm going off of the fact that most users make their RAID and have it storing the only copy of their data (i.e. not the backup). This is why people are so desperate to recover the array as opposed to simply restoring from backup.

    > If Drobo are using Caviar Greens, I would expect that their
    engineers have determined that their controller logic is not
    prone to dropping member HDDs, or simply programmed it that way.

    Yup. I consider a Drobo under the 'good' category, but it's external, and not in the same class as an internal (i.e. high speed) RAID HBA.

    > As for our experiments, I can definitely say that enabling polling
    on a Highpoint RocketRAID 2322 definitely causes a RAID 0 array
    to fail, quite often. As soon as I disabled that polling function,
    there have been no array failures -- NONE!
    > And, this difference occurred with Western Digital RAID Edition HDDs!!

    Not surprising. Highpoint has a long way to go on their firmware.

    > So, even with WD's TLER, a "bad" controller (using your terminology)
    can decide to drop a member disk and do so erroneously.

    Yup.

    > And, I don't think it's too much to ask that such polling be
    enabled in such a way that it occurs only when a given
    HDD is not already busy -OR- be configurable by allowing
    the User / System Administrator to disable polling entirely,
    as I have done with the RR2322.

    That's where Areca gets it right. Background tasks (rebuild, check, etc), are treated as just that. When idle (i.e. no requests from the host), an Areca will work at 100%. When the host requests IO, the card drops to the user selected priority (in %). Once idle again, the controller goes back to 100%. This juggling is handled very well as implemented by the Areca firmware.

    NCQ functions at the system/cache side as well as the HDD side, simultaneously. Caching on Areca cards is also incredible. I upgraded my card to 1GB of cache (upgradeable SO-DIMM). I went cheaper and used non-ECC RAM and it's been flawless. I beat on this array constantly (a few VM's, downloads constantly keeping the array active, bulk file copy from my OS SSD RAID to the array, etc), and it never so much as hiccups. The closest I've had to a dropout was after I shut the system down for maintenance. Upon restart the Areca alarmed and informed me a drive was missing from the array. The default is to keep the array offline in this situation, so after correcting a loose SATA Power adapter the array was back up without the need for a rebuild.
    Allyn Malventano, CTNC, USN
    PCPer Storage Editor

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •