Linux Archive

Linux Archive (http://www.linux-archive.org/)
-   Fedora User (http://www.linux-archive.org/fedora-user/)
-   -   Help Diagnose Slow Disc Access (http://www.linux-archive.org/fedora-user/324968-help-diagnose-slow-disc-access.html)

Mike McCarty 02-12-2010 11:04 AM

Help Diagnose Slow Disc Access
 
Alan Cox wrote:

[...]

> You also want to avoid two disks on one cable as the IDE interface only
> allows one of them to be active at a time so its a good way to cripple
> performance.

My previous reply may not have been quite motivational enough.

There is a physical constraint as to where the discs can be
mounted in the case which makes it difficult to run then on
separate controller interfaces. These discs are 3.5" form
factor, and the CD and DVD drives I have are 5.25" form factor.

The natural thing would be to put the two disc drives as the masters
of the two ATA channels provided, and make the two removable
medium drives be the slaves. However, the 5.25" bays are about
eight inches from the 3.5" bays, so this is physically difficult
to realize.

Mike
--
p="p=%c%s%c;main(){printf(p,34,p,34);}";main(){pri ntf(p,34,p,34);}
Oppose globalization and One World Governments like the UN.
This message made from 100% recycled bits.
You have found the bank of Larn.
I speak only for myself, and I am unanimous in that!
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines

Mikkel 02-12-2010 03:54 PM

Help Diagnose Slow Disc Access
 
On 02/12/2010 05:31 AM, Don Quixote de la Mancha wrote:
>
> A while back I downloaded all the diagnostics from all the drive
> vendors, and burned a CD for each one. I also keep them around on a
> filesystem where I archive all my software installers. They're good
> things to have on hand.
>
You may want to look at the Ultimate Boot CD. Less CDs to have to
worry about to get the same thing.

http://www.ultimatebootcd.com/

Mikkel
--

Registered Linux User #16148 (http://counter.li.org/)

--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines

Mikkel 02-12-2010 03:59 PM

Help Diagnose Slow Disc Access
 
On 02/12/2010 05:40 AM, Alan Cox wrote:
>
> You also want to avoid two disks on one cable as the IDE interface only
> allows one of them to be active at a time so its a good way to cripple
> performance.
>
Dumb question - what happens if you have a slow interface on the
CD/DVD drive? Does the driver change access modes depending on if it
is talking to the master or slave, or does it pick the lowest common
denominator? I seam to remember that being a problem at one time,
but things have changed a lot sense then...

Mikkel
--

Registered Linux User #16148 (http://counter.li.org/)

--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines

Alan Cox 02-12-2010 04:51 PM

Help Diagnose Slow Disc Access
 
On Fri, 12 Feb 2010 10:59:22 -0600
Mikkel <mikkel@infinity-ltd.com> wrote:

> On 02/12/2010 05:40 AM, Alan Cox wrote:
> >
> > You also want to avoid two disks on one cable as the IDE interface only
> > allows one of them to be active at a time so its a good way to cripple
> > performance.
> >
> Dumb question - what happens if you have a slow interface on the
> CD/DVD drive? Does the driver change access modes depending on if it
> is talking to the master or slave, or does it pick the lowest common
> denominator? I seam to remember that being a problem at one time,
> but things have changed a lot sense then...

Most controller hardware supports switching the mode according to the
device. In some of the other cases the drivers (particularly the libata
ones) also know how to change mode in software when switching device.
SATA killed that problem off.

Modern hardware with AHCI interfaces enabled dump all the work on the
controller and allow multiple commands to be queued at once. That
improves things even further.

Alan

--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines

Mike McCarty 02-19-2010 08:50 PM

Help Diagnose Slow Disc Access
 
Don Quixote de la Mancha wrote:
> All of the hard drive vendors provide disk drive diagnostic tools,
> that are able to access vendor-specific - and undocumented - firmware
> in their drives. This diagnostic firmware is able to diagnose drive
> hardware problems in a much more thorough way than the vendor-neutral
> S.M.A.R.T. is able to.
>
> These utilities are always provided in the form of DOS boot disk
> images; one generally has a choice of making a floppy or a CD-ROM.

I downloaded said software, and burnt a CD-ROM. I ran the diagnostics
on both discs (both are WDs, but of different sizes). The smaller
one passed both a "quick" test, and an "extended" full surface scan
test, and both in about the amount of time the tool estimated. The
larger one (the one I'm having performance problems with) failed the
"quick" test, due to timeout, after several times the estimated
run time, but passed the "extended" full surface scan, though it took
significantly longer than estimated. The estimated time was just over
15 hours, but the test ran 83 hours 33 minutes.

> Finally they all have a destructive test, in which the diagnostic
> writes zeroes to every sector of the drive.

I did not try to run the destructive tests. There is one which performs
a write test, and another which is not a test, but just intended to
write zeros to all sectors.

> No matter what, if you think one of your drives might be flaky, back
> them both up at once, before doing anything else.

That goes without saying.

> Being fully backed up also gives you the advantage that you can then
> run the destructive sector-zeroing test. I feel it's a good thing to
> do in any case, just to "exercise the bits".

I'm not prepared to run another 83 hours non stop off line.

[...]

> Hope That Help,

Well, so far what the software has told me is that the disc appears
to be OK, but very slow, which is what I already knew.

I want some help getting information out of the kernel to see what
it thinks. Anyone familiar with how to do that?

I've wondered whether DMA might be disabled, or perhaps it's not
running with interrupts, but hdparm seems to think that both drives
are essentially running the same...

(ok drive)

# hdparm /dev/hda

/dev/hda:
multcount = 16 (on)
IO_support = 1 (32-bit)
unmaskirq = 1 (on)
using_dma = 1 (on)
keepsettings = 0 (off)
readonly = 0 (off)
readahead = 256 (on)
geometry = 65535/16/63, sectors = 78165360, start = 0

(slow drive)

# hdparm /dev/hdb

/dev/hdb:
multcount = 16 (on)
IO_support = 1 (32-bit)
unmaskirq = 1 (on)
using_dma = 1 (on)
keepsettings = 0 (off)
readonly = 0 (off)
readahead = 256 (on)
geometry = 38913/255/63, sectors = 625142448, start = 0

It's odd that hdparm is unable to notice that the disc is slow.

Mike
--
p="p=%c%s%c;main(){printf(p,34,p,34);}";main(){pri ntf(p,34,p,34);}
Oppose globalization and One World Governments like the UN.
This message made from 100% recycled bits.
You have found the bank of Larn.
I speak only for myself, and I am unanimous in that!
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines

Mike McCarty 02-23-2010 03:51 AM

Help Diagnose Slow Disc Access
 
Mike McCarty wrote:
> My machine has been running slower and slower, and top seems
> to indicate lots of I/O wait. I have two ATA discs on a single
> cable, wired for cable select. The master is much faster than
> the slave, which seems to indicate a hardware, possibly disc,
> problem.

So, is there any way I can query the kernel's information about
the disc, and find out what it thinks may be the problem?

Mike
--
p="p=%c%s%c;main(){printf(p,34,p,34);}";main(){pri ntf(p,34,p,34);}
Oppose globalization and One World Governments like the UN.
This message made from 100% recycled bits.
You have found the bank of Larn.
I speak only for myself, and I am unanimous in that!
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines

Dave Stevens 02-23-2010 10:27 PM

Help Diagnose Slow Disc Access
 
On Friday 19 February 2010 13:50:59 Mike McCarty wrote:
> Don Quixote de la Mancha wrote:
> > All of the hard drive vendors provide disk drive diagnostic tools,
> > that are able to access vendor-specific - and undocumented - firmware
> > in their drives. This diagnostic firmware is able to diagnose drive
> > hardware problems in a much more thorough way than the vendor-neutral
> > S.M.A.R.T. is able to.
> >
> > These utilities are always provided in the form of DOS boot disk
> > images; one generally has a choice of making a floppy or a CD-ROM.
>
> I downloaded said software, and burnt a CD-ROM. I ran the diagnostics
> on both discs (both are WDs, but of different sizes). The smaller
> one passed both a "quick" test, and an "extended" full surface scan
> test, and both in about the amount of time the tool estimated. The
> larger one (the one I'm having performance problems with) failed the
> "quick" test, due to timeout, after several times the estimated
> run time, but passed the "extended" full surface scan, though it took
> significantly longer than estimated. The estimated time was just over
> 15 hours, but the test ran 83 hours 33 minutes.
>
> > Finally they all have a destructive test, in which the diagnostic
> > writes zeroes to every sector of the drive.
>
> I did not try to run the destructive tests. There is one which performs
> a write test, and another which is not a test, but just intended to
> write zeros to all sectors.
>
> > No matter what, if you think one of your drives might be flaky, back
> > them both up at once, before doing anything else.
>
> That goes without saying.
>
> > Being fully backed up also gives you the advantage that you can then
> > run the destructive sector-zeroing test. I feel it's a good thing to
> > do in any case, just to "exercise the bits".
>
> I'm not prepared to run another 83 hours non stop off line.
>
> [...]
>
> > Hope That Help,
>
> Well, so far what the software has told me is that the disc appears
> to be OK, but very slow, which is what I already knew.
>
> I want some help getting information out of the kernel to see what
> it thinks. Anyone familiar with how to do that?
>
> I've wondered whether DMA might be disabled, or perhaps it's not
> running with interrupts, but hdparm seems to think that both drives
> are essentially running the same...
>
> (ok drive)
>
> # hdparm /dev/hda
>
> /dev/hda:
> multcount = 16 (on)
> IO_support = 1 (32-bit)
> unmaskirq = 1 (on)
> using_dma = 1 (on)
> keepsettings = 0 (off)
> readonly = 0 (off)
> readahead = 256 (on)
> geometry = 65535/16/63, sectors = 78165360, start = 0
>
> (slow drive)
>
> # hdparm /dev/hdb
>
> /dev/hdb:
> multcount = 16 (on)
> IO_support = 1 (32-bit)
> unmaskirq = 1 (on)
> using_dma = 1 (on)
> keepsettings = 0 (off)
> readonly = 0 (off)
> readahead = 256 (on)
> geometry = 38913/255/63, sectors = 625142448, start = 0
>
> It's odd that hdparm is unable to notice that the disc is slow.

what does hdparm -tT /dev/hdb show?

Dave

>
> Mike
>
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines

Kam Leo 02-24-2010 01:35 AM

Help Diagnose Slow Disc Access
 
On Fri, Feb 19, 2010 at 1:50 PM, Mike McCarty
<Mike.McCarty@sbcglobal.net> wrote:
> Don Quixote de la Mancha wrote:

[snip]

> I downloaded said software, and burnt a CD-ROM. I ran the diagnostics
> on both discs (both are WDs, but of different sizes). The smaller
> one passed both a "quick" test, and an "extended" full surface scan
> test, and both in about the amount of time the tool estimated. The
> larger one (the one I'm having performance problems with) failed the
> "quick" test, due to timeout, after several times the estimated
> run time, but passed the "extended" full surface scan, though it took
> significantly longer than estimated. The estimated time was just over
> 15 hours, but the test ran 83 hours 33 minutes.

Failing the "quick" test and long completion times are sure signs that
the drive is in trouble. You can try reformatting the drive to see if
that improves performance (doubtful). Good luck finding a new IDE
drive. You might have to use a SATA drive with a SATA-IDE adapter or
buy a SATA controller and change all your hard drives to SATA.


[snip]
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines

Don Quixote de la Mancha 02-24-2010 04:29 AM

Help Diagnose Slow Disc Access
 
On Tue, Feb 23, 2010 at 6:35 PM, Kam Leo <kam.leo@gmail.com> wrote:
> Failing the "quick" test and long completion times are sure signs that
> the drive is in trouble. You can try reformatting the drive to see if
> that improves performance (doubtful). Good luck finding a new IDE
> drive. You might have to use a SATA drive with a SATA-IDE adapter or
> buy a SATA controller and change all your hard drives to SATA.

Both SATA controllers and SATA drives are cheap and easy to find, and
there are many, many advantages to SATA over parallel IDE.

Didn't you say that the placement of drives in your case was
constrained somehow? One of the most significant advantages of SATA -
in fact the main reason we now use so many kinds of serial busses now,
rather than the parallel busses of old, is that SATA cables can be
quite long, and are thin and flexible. So you could put SATA drives
anywhere you needed to and just get a long cable.

An internal SATA to external eSATA adapter only costs ten bucks. With
that and an eSATA enclosure for fifteen or twenty dollars, you could
put your drives entirely outside your computer case. That's what I
use for my two disk-to-disk backup drives: I leave one connected to
eSATA and the other in a bank safe deposit box. Once a week I swap
them. The bank is quite some ways from my home, so even a direct
nuclear strike on my home office would not cause the loss of more than
a week's worth of my work.

SATA CD and DVD burners are quite cheap now. I bought a Pioneer
Dual-Layer DVD burner for sixty bucks, that I think can now be had for
forty. I recently bought a SATA Plextor Blu-ray burner, that can burn
Dual-Layer Blu-rays that hold 50 GB. The unit only set me back I
think $230. I intend to use that to make Blu-ray copies of my
Disk-to-Disk backup drives, that will only require a few (but
expensive) disks to hold my entire career's worth of data.

A little over a year ago I bought four one-terabyte Western Digital
RE3 (RAID Edition) drives for the AMCC 3ware 9690 RAID 5 in my F11
box. Those were really high end drives, yet only cost $200 apiece.
Just your basic commodity drive, not a RAID Edition, would have
similar capacity for about half the price.

Finally SATA allows for better logical protocols, such as Tagged
Command Queueing, which allows multiple I/O transfers to be issued in
rapid succession, without having to wait for one to complete before
issuing the next. With parallel IDE, the entire bus is completely
tied up for the whole duration of a single command.

There are lots of SATA controllers that support Linux these days. A
two-port SATA Host Bus Adapter (HBA is the correct name, not
"controller") would cost maybe twenty bucks. I have a couple
four-port Promise SATA HBAs, and they have never given me any trouble.
The HBAs can have from two to eight or so ports, and may have some
internal ports and some external eSATA ports. eSATA is electrically
identical to internal SATA; it just uses a different connector than
the internal.

SATA allows such long cables because it uses twisted pairs for each
data channel, with differential signalling. Twisted pairs tend to
cancel out electrical interference all by themselves, whereas
differential signalling eliminates common-mode pickup. So a
differential twisted pair cable can be quite long, and used in very,
very electrically-noisy environments with no data transmission errors
at all.

Besides SATA, USB and FireWire use differential twisted pairs, as does
Serial Attached SCSI (SAS). SAS uses the same cabling as SATA does,
but with the SCSI logical protocol. The basic SATA uses the same
logical protocol as parallel IDE, so much of the industry's existing
investment into parallel IDE firmware and device drivers can still be
used for SATA - one reason that it's so cheap.

Please take my advice, my friend: the *last* thing you ever want to do
is store valuable data on a hard drive that you're not completely
confident of. While it is merely slow today, tomorrow it may start
irretrievably losing data. You might not be able to tell right away
that you have lost data, because you might not actually look at the
corrupted files until much later on.

(I've spent seven years doing all manner of storage programming. If
there is one thing I know about, it is failed hard drives.)

Don Quixote
--
Don Quixote de la Mancha
quixote@dulcineatech.com
http://www.dulcineatech.com

Dulcinea Technologies Corporation: Software of Elegance and Beauty.
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines

Les 02-24-2010 05:11 AM

Help Diagnose Slow Disc Access
 
On Tue, 2010-02-23 at 18:35 -0800, Kam Leo wrote:
> On Fri, Feb 19, 2010 at 1:50 PM, Mike McCarty
> <Mike.McCarty@sbcglobal.net> wrote:
> > Don Quixote de la Mancha wrote:
>
> [snip]
>
> > I downloaded said software, and burnt a CD-ROM. I ran the diagnostics
> > on both discs (both are WDs, but of different sizes). The smaller
> > one passed both a "quick" test, and an "extended" full surface scan
> > test, and both in about the amount of time the tool estimated. The
> > larger one (the one I'm having performance problems with) failed the
> > "quick" test, due to timeout, after several times the estimated
> > run time, but passed the "extended" full surface scan, though it took
> > significantly longer than estimated. The estimated time was just over
> > 15 hours, but the test ran 83 hours 33 minutes.
>
> Failing the "quick" test and long completion times are sure signs that
> the drive is in trouble. You can try reformatting the drive to see if
> that improves performance (doubtful). Good luck finding a new IDE
> drive. You might have to use a SATA drive with a SATA-IDE adapter or
> buy a SATA controller and change all your hard drives to SATA.
>
>
> [snip]
Microcenter had a good supply last time I was up there, and reasonably
priced, too. They are available on the web.

www.microcenter.com

Regards,
Les H

--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines


All times are GMT. The time now is 10:59 AM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.