Summary: Fujitsu 1.2G Disk

From: d3e101%simpsons.pnl.gov@pnlg.pnl.gov
Date: Fri Jan 25 1991 - 15:47:19 CST


I received 13 replies to my question on the problem I was haveing with Fujitsu M2266SA
drives on SS@'s running 4.1.1, and I solved the problem. First, here's the main section of
my question:

> Our site recently purchased 3 Fujitsu M2266SA 1.2G disks from Bridgeway for
>use in SS2's running 4.1.1. Each machine has the entire disk configured as
>one big /tmp partition, to use for computational chemistry applications. All
>3 disks had defect lists, so I newfs'ed and mounted them no problem.
>However, all 3 disks have several read hard errors (no write errors though)
>when the disk is used by these applications, or when I try to read large
>files off the disk. I find it highly unlikely that all 3 disks are faulty,
>and I even reformatted one of them after this happened, with no change. Here
>is a typical error:
>
>Jan 22 14:30:05 vulcan vmunix: sd2e: Error for command 'read'
>Jan 22 14:30:05 vulcan vmunix: sd2e: Error Level: Fatal
>Jan 22 14:30:05 vulcan vmunix: sd2e: Block 124624, Absolute Block: 124624
>Jan 22 14:30:05 vulcan vmunix: sd2e: Sense Key: Hardware Error
>Jan 22 14:30:05 vulcan vmunix: sd2e: Vendor 'FUJITSU' error code: 0x44

I also forgot to say in my message that the read errors were occuring on random blocks, during
heavy disk activity.

People suggested relabelling the disk, turning off synchronous SCSI, media errors, using
smaller partitions (unacceptable here), checking SCSI termination, SCSI ID, SCSI fuse, and
cable length, driver bugs, using it as tmpfs, and using a different shoebox. Most of these
were not applicable to my situation. However, the winning suggestion was to disable the
read-ahead cache on the disk. This seemed at the time to make sense, since there only
read errors, no write errors. After looking at the Fujitsu BBS, I found what seemed to be
a description of the jumper settings for the disk in files M2266S_H.JMP and M226XS_H.JMP.
It was also mentioned elsewhere on the BBS that some controllers have problems with read-ahead
cache enabled. It says the read-enable cache jumper is on the back of the disk, where you set
the SCSI ID, and is the 3rd one from the left. Take that jumper out and you're all set. I
don't see a performance decrease after doing this.

I also reformatted the disk according to sun-managers postings, but I don't know if it made any
difference. I originally didn't have to format the disk when it came out of the box,
but the defect list was not recognized when I used the new format.dat entry. Another thing I
noticed is the "show" command in format does not work. It claims the MD81 controller does not
support it, but other disks with this controller do support it. Finally, Bridgeway did not
supply the "Fujitsu Tech Tips".

Thanks to the following people:

  Ric Anderson <ric@cs.arizona.edu>
  merlyn@iWarp.intel.com
  Matt Crawford <matt@oddjob.uchicago.edu>
  markm%bit.uucp@BCLVXF
  shj@ultra.com
  guyton%condor@rand.org
  soleil!grina@rutgers.edu
* sgf@cfm.brown.edu
  sitongia@hao.ucar.edu
  aggarwal@nisc.jvnc.net
  edsr!tantalum!cheeks@uunet.UU.NET
  eap@bu-pub.bu.edu (Eric A Pearce)

  (Thanks to "John A" too)

* = the right answer

                        Bill Eggers
                        Molecular Science Research Center
                        Pacific Northwest Labs, Richland, WA
                        d3e101@pnl.gov



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:10 CDT