SCSI drive reliability problem

SCSI drive reliability problem

Post by Hann » Mon, 28 Jan 2002 06:32:38



I have now had three Fujitsu Ultra 160 SCSI drives failed on different
Linux installations.
The first two failed under a RH7.1 kernel 2.4.x system, with standard
and upgraded kernels, third with a AMD 1600 system running Mandrake 8.1
kernel 2.4.8. The RH box had a Asus built-in motherboard SCSI
controller, the Mandrake box an Adaptec 29160.
All these failures are almost identical. The box is running fine for
weeks, and reboots back just fine. After about 5-20 reboots, it just
wont' boot any more: "cant find partition table". If you reinstall, it
will work for a while again, then after couple weeks same thing happens
again.
I was about to build a new production server using these drives, now I
don't think that will be the case.
Any ideas?
Hannu
 
 
 

SCSI drive reliability problem

Post by Peter T. Breue » Mon, 28 Jan 2002 10:35:12



> I have now had three Fujitsu Ultra 160 SCSI drives failed on different
> Linux installations.
> The first two failed under a RH7.1 kernel 2.4.x system, with standard
> and upgraded kernels, third with a AMD 1600 system running Mandrake 8.1
> kernel 2.4.8. The RH box had a Asus built-in motherboard SCSI
> controller, the Mandrake box an Adaptec 29160.

I am running several (intel cpu) systems with 19160 and 29160
controller. My own systems have WD drives. The others are IBM and
fujitsu.

I also have some asus (dual) scsi mobos amongst those.

Quote:> All these failures are almost identical. The box is running fine for
> weeks, and reboots back just fine. After about 5-20 reboots, it just
> wont' boot any more: "cant find partition table". If you reinstall, it
> will work for a while again, then after couple weeks same thing happens
> again.

I have problems with the 29160 and the ibm - spontaneous reboots.
Possibly bad termination. No losing the partition table, but I have
seen this several times over the years with several combinations of
scsi.

Quote:> I was about to build a new production server using these drives, now I
> don't think that will be the case.
> Any ideas?

You need more data.

Peter

 
 
 

SCSI drive reliability problem

Post by Hann » Sat, 02 Feb 2002 06:08:20


Quote:> What chip?

The one in my dual processor box (motherboard Asus CUV4X-DLS) is LS!
53C1010-33 Ultra160/Ultra3

The other on is the Adaptec 29160.

Quote:> Are you using the "new" adaptec driver (aic7xxx.o)? It seems to be
> extremely buggy. I'm not even able to install my adaptec based system.
> It keeps putting something like "attempting to queue an abort message"
> in the onscreen syslog as soon as the installation starts to create the
> first filesystem.

Yes, first I had aic7xxx.o loaded as module, didn't work at all.
Then I upgraded the system to run kernel 2.4.17 and compiled the SCSI
driver aic7xxx and other Adaptec drivers into the kernel.

Quote:

> Since I ran into SMP issues I haven't used 2.4.x for quite some time -
> perhaps the Adaptec problem has been fixed.

Yes, there were a lot of SMP issues with the early 2.4 kernels. The new
one (2.4.17) appears to have some kind of support for the apic functions
of the new boards. Remains to be seen if the adaptec problem is still
there.

Quote:> Doesn't sound like a drive failure at all. Something seems to be
> overwriting the MBR. Might even be a boot sector virus.

No, the boot sector was fine. Linux booted and ran to the point were it
would mount sda1 (root filesystem)

Quote:> If you want to be able to buy the same type of drive a a replacement
> part better check if they still make it! Fujitsu has stopped making
> "desktop" drives some time ago. I'm not sure if that includes the SCSI
> drives.

I think they still make them. I got one as a replacement, slightly newer
model number. However, I may switch if I have another crash. Which Ultra
160 drives would you recommend?

Hannu

Quote:

> Ciao,
> Andreas

 
 
 

SCSI drive reliability problem

Post by Hann » Sat, 02 Feb 2002 10:52:27


Andreas;
Interesting comments, my experiences are very similar. We have been using
Fujitsu SCSI Ultra on Adaptec controllers for years, never had a problem.
A common factor: All those systems have been running linux 2.2.x kernels,
right now I have three going, one has uptime or 13 months, another one 8
months.
All these problems I mentioned have been under 2.4.x kernels, so I have
been very hesitant to put 2.4 into production.
I will let you know how my latest installation (2.4.17) goes.
Thanks
Hannu


>>>What chip?
>>The one in my dual processor box (motherboard Asus CUV4X-DLS) is LS!
>>53C1010-33 Ultra160/Ultra3

> IMHO NCR/Symbis/LSI is the best choice for Linux. The only "issue" i've
> ever run into with them is the lack of support for the LED-Out on Tekram
> cards. I bought some LED's and connected them directly to the drives.

> [Problems with "new" adaptec driver]
>>Yes, first I had aic7xxx.o loaded as module, didn't work at all.

> The aic7xxx_old module should at least get the job done. It may be a
> little slow.

>>Then I upgraded the system to run kernel 2.4.17 and compiled the SCSI
>>driver aic7xxx and other Adaptec drivers into the kernel.

> I prefer to load SCSI drivers as modules from an initrd.

>>>Since I ran into SMP issues I haven't used 2.4.x for quite some time -
>>>perhaps the Adaptec problem has been fixed.
>>Yes, there were a lot of SMP issues with the early 2.4 kernels. The new
>>one (2.4.17) appears to have some kind of support for the apic functions
>>of the new boards.

> It's a rather old Gigabyte 586DX(*) that works like a charm with 2.2
> kernels. I usually reach several weeks of uptime, ended by some power
> failure.

> *) Dual Socket-7, Intel 430HX Chipset, Onboard aic7880 UW-SCSI

>>Remains to be seen if the adaptec problem is still there.

> I'd be pleased if you could inform me of what you find out. I'd rather
> use the same kernel on all of my systems and I haven't managed to patch
> a 2.2 kernel with  support for LVM and the NIC built in SiS 735
> chipsets.

>>> Doesn't sound like a drive failure at all. Something seems to be
>>> overwriting the MBR. Might even be a boot sector virus.
>>No, the boot sector was fine. Linux booted and ran to the point were it
>>would mount sda1 (root filesystem)

> Absence of evidence is not evidence of absence. A lot of boot sector
> viruses save the sector's original contents somewhere else and load them
> from there, pretending everything's alright.

>>>If you want to be able to buy the same type of drive a a replacement
>>>part better check if they still make it! Fujitsu has stopped making
>>>"desktop" drives some time ago. I'm not sure if that includes the SCSI
>>>drives.
>>I think they still make them. I got one as a replacement, slightly newer
>>model number.

> Even if they don't make it any more they will keep drives on stock as
> replacements.

> Personally I wouldn't mind replacing a failed drive with a different
> model. I just know some companies with really strict policies. Replacing
> a drive with a different model could require extensive testing and
> returning a failed drive may not be allowed.

>>However, I may switch if I have another crash. Which Ultra 160 drives
>>would you recommend?

> It's a major problem to buy SCSI drives from anyone but IBM in this
> area. My DDYS seems OK to me. It's very fast and I was suprised that
> it's not nearly as loud as I had imagined.

> Ciao,
> Andreas

 
 
 

1. Replaced scsi hard drive can no longer find scsi tape drive

My Linux system had a hard disk crash just recently and I had to
replace the hard drive.  I did so and installed a "brand-new" copy of
RH 7.2 on it.  Previously it was running RH 6.2, but since I needed to
catch up a bit, it seemed like now was as good a time as any.....
The system is now up and running.

Now, I would like to restore my backup tapes on to it.  The system has
a HP SCSI DDS-3 tape drive on it, which I have used with tar to keep
backups.

The problem is Linux is no longer "seeing" my tape drive.

So, I look for the device file, and it looks "ok"

And I look to see if the device is known:
Ooops, no tape drive, just a disk.  

Am I missing support in the kernel:
...
Well, the kernel seems to know about scsi tapes.

I also checked the physical connections to the drive and they seem "ok".

I'm not certain what to check next.

Suggestions would be appreciated.

Thanks,
-Chris

*****************************************************************************

Compiler Resources, Inc.       Web Site   :  http://world.std.com/~compres  
3 Proctor Street               voice      :  (508) 435-5016
Hopkinton, MA  01748  USA      fax        :  (508) 435-4847  (24 hours)
------------------------------------------------------------------------------

2. Just a few questions before installation

3. Mount SCSI NCR 810 + ZIP DRIVE + SCSI DRIVE

4. Avaya wireless --> 3com NIC gateway/router setup questions.

5. Adding a SCSI controller changing boot drive to SCSI drive

6. missing files

7. Problems using cdrecord to scsi cdwriter from scsi hard drive

8. Help with strange message about "giant packet"?

9. SCSI Zip drive w/ Zip zoom scsi adapter problems...

10. SCSI problems Iomega Adaptec 152x SCSI Zip drive

11. reliability of CAM scsi driver on ASUS P2B-S?

12. Reliability of Backpack CD-ROM Drive?

13. Tape drive problem/SCSI problem?