[GLLUG] Disk drive controller question

[Richard Houser]

And turn on SMART!

2009/2/6 Bill Bartilson <bbartilson at comcast.net>

> Nope.  Not accounted for.  We have to copy a partition from B to A from
> each controller.  Could be some low level problem with the drive, how the
> partition table worked out on one vs. the other, or how you've got one of
> the cables routed when it gets hooked to the other drive.
> You're right that caching really should not be an issue over the period of
> sustained read/write we're discussing.
>
> As a tech guy for many years, when I see hoof prints, I think horses, not
> zebras.  It is very likely that you are dealing with something mundane and
> ordinary. (like a drive that ain't happy for some reason, or one
> controller's immunity OR sensitivity to some kind of noise) as opposed to
> something really out of the ordinary.  (like cosmic rays or other stellar
> phenomena)  Something that's unusual doesn't necessarily mean genuinely
> *weird*.
>
> Statistics aren't worth much by themselves.  Statistics based on my own
> experience, I tend to believe.  And somewhere between 80% and 90% of what I
> see are relatively simple problems.  Good troubleshooting will ALWAYS get
> you to the source of the problem.  You have yet to do more than half the
> work for eliminating the drives themselves as an issue.  When PC's (of any
> variety) go funky or act unusually, the vast majority have issues with power
> supply or disk drives.  Period.  ALL of the rest of the problems account for
> only 1 or 2 out of 10 repairs.
>
> Again, it's a matter of whether you actually care to find out or not.   I
> merely suggested possibilities for elimination.  You're welcome to try or
> not at your own discretion.  I heartily endorse Spinrite for anyone to use
> on any system.  It doesn't replace backup, but it obviates the need for
> disaster recovery much of the time.
>
> -B
>
>
> On Feb 6, 2009, at 4:41 PM, Stanley C. Mortel wrote:
>
>  Points 1 and 2 have already been accounted for.  Copy partition on A
> attached to on-board IDE to B attached to 3WARE: takes 5 minutes.  Now just
> disconnect the cables at the drives and swap them around.  Copy partition on
> A attached to 3WARE to B attached to on-board IDE:  takes 2 hours.  Point 3
> is, I suppose, possible, but the on-board IDE controller works just fine and
> isn't THAT old.  If it could only write about 1/24th as fast as the 3WARE
> card then either the system would suck in normal operation or the 3WARE card
> would be worth a lot of money.  Besides that, the on-board IDE controller
> would have to be only writing at 1/24th of what the drive is capable of.  I
> haven't taken a look at the specs of the drive, but, again, the system and
> IDE controller are not that old, and they work as expected.  Also, they are
> both on the same PCI-33 bus.  I may be that the 3WARE card is caching a
> bunch, but for a sustained transfer like this, that wouldn't account for the
> difference.
>
> I think.
>
> Stan
>
>
>
> Bill Bartilson wrote:
>
> One is obviously faster than the other.   :)
>
> All kidding aside, if you want to shed light on what the actual
> differences are:
>
> 1.  Swap the drives.  I know you say they are identical, but they
> can't be.  Each drive has different physical parts.  One drive might
> be dying, one might be healthy.  At any rate, it's only science if you
> can duplicate it.  Mark one 'A' and one 'B' and try each on each
> different controller, copying each way, and note any differences, and
> whether the differences follow the drive or not.
>
> 2.  Repeat the same deal with each other item.  (IDE wires, different
> jumper settings on the drive, anything else you can think of to
> change.)  Keep good notes.
>
> 3.  Try a different piece of software (Gparted or something)  Perhaps
> the code written for reading data with controller X works much better
> with controller X, but only for long reads, and works like crap for
> long writes.  I dunno.
>
> Via process of elimination, you'll discover that one controller reads
> or writes a lot faster, one of the drives is actually not so good, or
> something in the code itself is different.
>
> This all presumes it's not merely an intellectual question and you
> care to 'know' the answer.  I for one would be interested to hear your
> findings if you care to try.
>
> Me thinks the most likely scenario is that one of the drives is
> getting funky.  In which case I'd say Spinrite them both before you
> try anything else.
>
> Regards,
> B
>
>
> On Feb 6, 2009, at 2:35 PM, Stanley C. Mortel wrote:
>
>
>
>  OK, here's the deal.  I'm using Acronis to copy a partition from one
> IDE
> drive to another identical drive, both are 120 GB with a single
> partition.  One drive is connected to the on-board IDE controller, the
> other is connected to a 3WARE RAID controller with only one drive on
> it.  Each drive is alone on its channel / cable, set to cable
> select, on
> the right connector, etc.  If I copy from the IDE to the 3WARE, it
> takes
> about 5 minutes.  If I try it the other way around, it tells me it is
> going to take 2 hours.  There is nothing in my knowledge base to
> explain
> this.  Anyone know why this would happen?  If you don't know for sure,
> any reasonable ideas?
>
> Stan
>
>
> _______________________________________________
> linux-user mailing listlinux-user at egr.msu.eduhttp://mailman.egr.msu.edu/mailman/listinfo/linux-user
>
>  _______________________________________________
> linux-user mailing listlinux-user at egr.msu.eduhttp://mailman.egr.msu.edu/mailman/listinfo/linux-user
>
>
>
> _______________________________________________
> linux-user mailing list
> linux-user at egr.msu.edu
> http://mailman.egr.msu.edu/mailman/listinfo/linux-user
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.egr.msu.edu/mailman/public/linux-user/attachments/20090207/59600483/attachment.html