rtidd@mwunix.mitre.org (Randall Tidd) (05/30/90)
Following is an in-depth description of some {hardware/software}?
problems we have been having with our Sun 3/260. If you don't care, just
skip this message.
Our configuration is as follows:
Sun 3/260, 16 megs RAM
Boot PROM 2.7
3 Maxtor 8760-E 600meg hard drives (SCSI)
SCSI-3 Controller
ALM-II Multiport Board
Parallax Graphics card
The system ran for 18 months with no problems under SunOS 3.5.1. After
upgrading to 4.0.3 at the beginning of April, we have been having a series
of strange file system errors including: duplicate inodes, unassigned
blocks, corrupted files, etc. The most common error seems to be an
"unexpected inconsistency error" from fsck, causing us to fun fsck
manually and seeing a list of files that were zorched. Corrupt files are
common as well, such as bad binaries or files with 000 permissions, 0
bytes and file type zero.
These problems happen over all hard disks, on any partition, seemingly
without pattern (at least not that i've seen in 6 weeks). Even in short
sessions, the next reboot of the machine often has many suprises from
fsck.
What we've done:
All hardware diagnostics checked out OK. We got Sun hardware support to
come out and do the diags, and they checked out OK. The errors happen
even under the GENERIC kernel. We re-installed the OS (several times),
including formatting the drives and reinstalling from tape. We got support
from Sun {hard,soft}ware, Emulex, Maxtor and Aretecon. Sun hardware
couldn't find anything wrong, the rest of them couldn't fix it. We
swapped out the CPU board with no effect.
The Artecon guy (who was quite competent) took away our Artecon disk
subsystem (with backplane and all cabels) and shipped it off; that was a
few days ago. We are now running an Artecon shoebox with only 2 of the 3
drives in it now, and still getting the errors.
I think that's it. I've tried everything I know, and everything anyone
else can think of. I'm posting this as sort of a last resort, before we
get rid of the Sun and either buy another one or buy something else
entirely. If anyone has any info or any suggestions, I would appreciate
it very much.
Thanks.
Randy Tidd
rtidd@mwunix.mitre.org