RE: disk problems or false alarm??

[Guolin Cheng]

Hi, Alexander,

 Thanks for your reply.

 But in fact, I'm using Maxtor disk diagnosis 4.06 diskette offline for long time. I
use it for drive connection test, and quick (90seconds) test. Then I  boot the Linux
failed machines into netRepair mode, to run "badblocks" programs on all 4 hard
disks (/dev/hda, /dev/hdb, /dev/hdc, /dev/hdd) one by one to detect sector errors.

My only concern is, I have been using "hdparm -d1 -c3 -m16 -a16 -A1 -u1 -W1 -k1
-K1" command on all 4 PATA hard drives, to speed up disk access speed, and improve
machines' responsiveness. All other options seems OK except "-u1", which,
according to manual, may bring "massive filesystem corruption" (Although for 3
years I have seen no file system corruptions because of that). But if I don't enable
the options, the Linux boxes will response way slow to keyboard when high-speed data
transfer happens.

 Any suggestions or hints are greatly appreciated.

 Thanks.
 --Guolin Cheng




-----Original Message-----
From: Alexander Dalloz [mailto:[e-mail]] 
Sent: Friday, April 30, 2004 12:46 PM
To: For users of Fedora Core releases
Subject: RE: disk problems or false alarm??

Am Fr, den 30.04.2004 schrieb Guolin Cheng um 21:01:

>  Hi, jludwig,
>  
>   Thanks for your helpful information.
>  
>   Because I'm running Linux, so I assume there are no viruses. Then comes
>  several questions:
>  
>  1, How can I know whether all the spare sectors are in use and the disk
>  will lose data, or it is just the beginning of disk failure?
>  
>  2, How I can identify that the hard drive becomes dying at the first
>  minute? 
Use the drive sanity check tool by the drive manufacturer. Hitachi/IBM,
Maxtor, WD, they all have such a tool. 

>  3, How to identify the malfunctioning hard drives? Should I idle the
>  machine and test hard drives one by one to figure it out? Mostly it is
>  the faiure-reporting hard drive failed, but I remember for sure, in a
>  few cases, other alternative hard drives failed instead.  
hda = master on primary controller
hdb = slave on primary controller
hdc = master on secondary controller
...

>  4, Should I replace hard drives when I first see this kind of disk
> error
>  messages in case data begin to lose? 
First check its state with a tool and other facts like cables. Very old
hard drives can only have problems with DMA. You may use hdparm to check
drive's setup.

>   Thanks a LOT...
>  
>  Guolin Cheng 
Alexander


-- 
Alexander Dalloz | Enger, Germany | GPG key 1024D/ED695653 1999-07-13
Fedora GNU/Linux Core 1 (Yarrow) on Athlon CPU kernel 2.4.22-1.2188.nptl
Sirendipity 21:41:36 up 3 days, 20:30, load average: 0.45, 0.64, 0.66 
                   [ Γνωθι σ'αυτον - gnothi seauton ]
             my life is a planetarium - and you are the stars

disk problems or false alarm??

[Alexander Dalloz]

Am Fr, den 30.04.2004 schrieb Guolin Cheng um 22:54:


>  My only concern is, I have been using "hdparm -d1 -c3 -m16 -a16
> -A1 -u1 -W1 -k1 -K1
>  " command on all 4 PATA hard drives, to speed up disk access speed, and improve
>  machines' responsiveness. All other options seems OK except "-u1", which,
> according
>  to manual, may bring "massive filesystem corruption" (Although for 3 years I
> have seen
>  no file system corruptions because of that). But if I don't enable the options, the
> Linux
>  boxes will response way slow to keyboard when high-speed data transfer happens. 
>   Guolin Cheng 
Well, forcing such agressive settings like you did is often cause for
trouble. I don't wonder any more. You did not mention such
non-selfdetected settings in your first mail. Communication between the
hard drive and the motherboard hardware using the chipset specific
driver is critical. In most every case you should let the kernel
autodetect the drives settings and not force things.

Alexander


-- 
Alexander Dalloz | Enger, Germany | GPG key 1024D/ED695653 1999-07-13
Fedora GNU/Linux Core 1 (Yarrow) on Athlon CPU kernel 2.4.22-1.2188.nptl
Sirendipity 23:09:23 up 3 days, 21:58, load average: 0.29, 0.18, 0.18 
                   [ Γνωθι σ'αυτον - gnothi seauton ]
             my life is a planetarium - and you are the stars

disk problems or false alarm??

[jludwig]

On Fri, 2004-04-30 at 17:14, Alexander Dalloz wrote:
>  Am Fr, den 30.04.2004 schrieb Guolin Cheng um 22:54:
>  
>  
>  > My only concern is, I have been using "hdparm -d1 -c3 -m16 -a16 -A1 -u1 -W1 -k1
> -K1
>  > " command on all 4 PATA hard drives, to speed up disk access speed, and
> improve
>  > machines' responsiveness. All other options seems OK except "-u1",
> which, according
>  > to manual, may bring "massive filesystem corruption" (Although for 3 years
> I have seen
>  > no file system corruptions because of that). But if I don't enable the options,
> the Linux
>  > boxes will response way slow to keyboard when high-speed data transfer happens.
>  
>  >  Guolin Cheng
>  
>  Well, forcing such agressive settings like you did is often cause for
>  trouble. I don't wonder any more. You did not mention such
>  non-selfdetected settings in your first mail. Communication between the
>  hard drive and the motherboard hardware using the chipset specific
>  driver is critical. In most every case you should let the kernel
>  autodetect the drives settings and not force things.
>  
>  Alexander 
I might add that I see nothing about bus speed. The default for linux
has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
speed).
I don't know about your system's maximum bus speed or your drives.
-- 
jludwig <wralphie(a)comcast.net&gt;

disk problems or false alarm??

[William Hooper]

jludwig said:
>  I might add that I see nothing about bus speed. The default for
> linux
>  has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
>  speed). 
I highly doubt you have a 133MHz PCI bus.  I also doubt you have tested to
verify this "quadrupule" drive speed.

-- 
William Hooper

disk problems or false alarm??

[jludwig]

On Fri, 2004-04-30 at 18:28, William Hooper wrote:
>  jludwig said:
>  > I might add that I see nothing about bus speed. The default for linux
>  > has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
>  > speed).
>  
>  I highly doubt you have a 133MHz PCI bus.  I also doubt you have tested to
>  verify this "quadrupule" drive speed.
>  
>  -- 
>  William Hooper 
Correct, newer systems do support this speed. Also the hard
drive(s)
must be considered in this issue.
-- 
jludwig <wralphie(a)comcast.net&gt;

disk problems or false alarm??

[William Hooper]

jludwig said:
>  On Fri, 2004-04-30 at 18:28, William Hooper wrote:
> > jludwig said:
> > > I might add that I see nothing about bus speed. The default for linux
> > > has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
> > > speed).
> >
> > I highly doubt you have a 133MHz PCI bus.  I also doubt you have tested
> > to
> > verify this "quadrupule" drive speed.
> >
> > --
> > William Hooper
>
>  Correct, newer systems do support this speed. 
Not unless you have PCI-X and a PCI-X hard drive controller.

>  Also the hard drive(s)
>  must be considered in this issue. 
Hard drives have nothing to do with the PCI bus speed.

-- 
William Hooper

disk problems or false alarm??

[jludwig]

On Fri, 2004-04-30 at 18:50, William Hooper wrote:
>  jludwig said:
>  > On Fri, 2004-04-30 at 18:28, William Hooper wrote:
>  >> jludwig said:
>  >> > I might add that I see nothing about bus speed. The default for linux
>  >> > has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
>  >> > speed).
>  >>
>  >> I highly doubt you have a 133MHz PCI bus.  I also doubt you have tested
>  >> to
>  >> verify this "quadrupule" drive speed.
>  >>
>  >> --
>  >> William Hooper
>  >
>  > Correct, newer systems do support this speed.
>  
>  Not unless you have PCI-X and a PCI-X hard drive controller.
>  
>  > Also the hard drive(s)
>  > must be considered in this issue.
>  
>  Hard drives have nothing to do with the PCI bus speed.
>  
>  -- 
>  William Hooper 
I see the error here. We were talking about hard drive/IDE busses
and
somehow PCI slipped in.-- 
jludwig <wralphie(a)comcast.net&gt;

disk problems or false alarm??

[William Hooper]

jludwig said:

>  I see the error here. We were talking about hard drive/IDE busses
> and
>  somehow PCI slipped in.-- 
No, idebus is PCI speed.

-- 
William Hooper

disk problems or false alarm??

[Alexander Dalloz]

Am Sa, den 01.05.2004 schrieb jludwig um 00:32:

>  > > I might add that I see nothing about bus speed. The default
> for linux
>  > > has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
>  > > speed).
>  > 
>  > I highly doubt you have a 133MHz PCI bus.  I also doubt you have tested to
>  > verify this "quadrupule" drive speed.
>  > 
>  > William Hooper
>  Correct, newer systems do support this speed. Also the hard drive(s)
>  must be considered in this issue. 
>  jludwig <wralphie(a)comcast.net&gt; 
Be aware, that idebus does not mean UDMA speed (like UDMA33 or UDMA100)
or the speed in MB/sec. On actual PC architecture the 32bit PCI bus is
rated at 33 MHz as well the IDE bus. So setting idebus to anything other
than the default of 33 is nonsense.

Alexander


-- 
Alexander Dalloz | Enger, Germany | GPG key 1024D/ED695653 1999-07-13
Fedora GNU/Linux Core 1 (Yarrow) on Athlon CPU kernel 2.4.22-1.2188.nptl
Sirendipity 00:46:43 up 3 days, 23:35, load average: 0.25, 0.20, 0.17 
                   [ Γνωθι σ'αυτον - gnothi seauton ]
             my life is a planetarium - and you are the stars

disk problems or false alarm??

[jludwig]

On Fri, 2004-04-30 at 18:51, Alexander Dalloz wrote:
>  Am Sa, den 01.05.2004 schrieb jludwig um 00:32:
>  
>  > > > I might add that I see nothing about bus speed. The default for linux
>  > > > has been 33Mhz for IDE systems (adding idebus=133 quadruples drive
>  > > > speed).
>  > > 
>  > > I highly doubt you have a 133MHz PCI bus.  I also doubt you have tested to
>  > > verify this "quadrupule" drive speed.
>  > > 
>  > > William Hooper
>  > Correct, newer systems do support this speed. Also the hard drive(s)
>  > must be considered in this issue.
>  
>  > jludwig <wralphie(a)comcast.net&gt;
>  
>  Be aware, that idebus does not mean UDMA speed (like UDMA33 or UDMA100)
>  or the speed in MB/sec. On actual PC architecture the 32bit PCI bus is
>  rated at 33 MHz as well the IDE bus. So setting idebus to anything other
>  than the default of 33 is nonsense.
>  
>  Alexander 
I stand corrected -- must be sleeping today!
-- 
jludwig <wralphie(a)comcast.net&gt;