Nope, not solved. still plenty of errors with the Adaptec SCSI RAID card, even using the older firmware. I did quite a few test writes to the RAID array, which were all fine. I then copied the kernel source from the /usr/src/linux to the RAID array and it locked up.

I now have the fun of dealing with Adaptec Technical Support :)

We have been plagued with same/similar problems on multiple servers using aacraid drivers. This random I/O error crash/reboot cycle has forced us to migrate towards LSI raid controllers. I haven’t had a single hiccup with the megaraid driver yet. We tried all the things people have mentioned regarding disabling ACPI, disabled Hyperthreading etc.. etc.. Nothing. Look familiar?

After this, just a continual scroll of I/O errors.....

For reference: Adaptec 2120S 64 bit w/ latest firmware ,u320 seagate w/ latest drive firmware, dual 2.4ghz xeon, intel SE7501CW2 board, sc5250 chasis w/ Intel hotswap. Have talked to both Seagate and Adaptec...no progress after a years of correspondence.

Oh btw, Adaptec Firmware is B6011

#!/bin/sh

total=1

while [ "$total" -le 20 ]
do
date >> /root/io_test.log
rm -rf /data1/*
count=1
while [ "$count" -le "40" ]
do
cp -r /usr/src/linux-2.6.4-52 /data1/thing$count
count=$((count+1))
done
total=$((total+1))
done

This tends to zap the aacraid module after about 10 minutes every time.

ya there is a aacraid module in 2.6.9 (part of source tree) i'm have the same problems w/ it..

recently, we just purchased a 2410SA sata raid adapter and after a few mins of being up and in use i get the same errors:

# dmesg
Buffer I/O error on device sda3, logical block 9
lost page write due to I/O error on sda3
scsi0 (0:0): rejecting I/O to offline device
...

from the looks of it this seems to be a problem w/ the aacraid driver (since the rest of you have similar or the same problems)

anyone try compiling the 1.1.5-2326 (off adaptecs site) against 2.6.9
?

In file included from /root/aacraid/linit.c:55:
drivers/scsi/hosts.h:1:2: warning: #warning "This file is obsolete, please use instead"
/root/aacraid/linit.c: In function `aac_eh_reset':
/root/aacraid/linit.c:898: warning: implicit declaration of function `scsi_sleep'
/root/aacraid/linit.c: At top level:
/root/aacraid/linit.c:1041: error: unknown field `highmem_io' specified in initializer
make[1]: *** [/root/aacraid/linit.o] Error 1
make: *** [_module_/root/aacraid] Error 2

i cant even compile it against 2.6.8* different errors however - mostly syntax errors (compiling wrong?)

i hope there's a solution soon i have 4 300 gig sata drives just sitting there collecting dust! =(

- dan

Yes, I would like to know what aacraid source it was that worked.
I tried compiling the 1.1.5-2326 from Adaptec against a Suse9.1 kernel and it failed with the same problem.

Since I posted some time ago (the BIG error posts at the top) the Adaptec 2120S RAID card broke. Returned to Adaptec, who then sent us a new one.
All plugged in and still the same I/O problems.
I am now going to add to my open technical support query with Adaptec and ask them for a new card/code that works.

PERFORMANCE
===========
The whole reason we wanted to use Suse9.1 in the first place was that the performance of the 2.4 kernel with a continuous write going on was absolutely awful. We got the 2120S up to 15MB/sec, but interactive login was not interactive at all.
Suse9.1 Pro manages 40MB/sec, but is rather unstable when doing multiple file operations (takes about 10 minutes to kill and is very repeatable).

The simple performance test we use is:
time dd if=/dev/zero of=./8gb bs=1024k count=8192

So I put Windows XP on the intel box. Using Cygwin dd the write test gave 20MB/sec, although how direct this is to the hardware I do not know.

A freebie performance tester for HDDs gives 40MB/sec under windows for continuous writes to our RAID5 array. Gives 80MB/sec for bursts and a read of 115MB/sec

Therefore it does look like the accraid driver gives pretty good performance (although we are still not impressed with the card itself), except for the stability issues.

Other people on this thread have mentioned megaraid and LSI - it would be really helpful if they could post any performance data.
i.e. please run:
time dd if=/dev/zero of=./8gb bs=1024k count=8192

with an LSI card and the megraid thang. We would be very interested if you get 60MB/sec of above.

cheers.

I am using the firmware version 7244 on my 2410SA..

i was readong on linux-scsi kern dev list and saw that it's possible the firmware on the drive might be the issue (huh?)

i'm using:

A0391552 300GB 7200RPM DiamondMax 10 Serial ATA Internal Hard Drive

anyone else having problems w/ diamond max and adaptec sata raid?
maxtor has no updates for firmware on the sata drive (?)

our raid setup is 4 diamond max drives on a raid 0

after fdisk'ing and formating partitions it dies with the scsi hang error message... after a hardboot the adaptec BIOS shows the drives as 150 GB drives instead of 300GB they showed originally (weird?).. i then have to rebuild the array.... to show the drives as 300GB and start over..

anyone else having these sort of problems in addition to the aacraid driver saying their adaptec is hanging?

-daniel

syslog stuff (again) :
Nov 5 06:20:22 (none) user.err kernel: aacraid: Host adapter reset request. SCSI hang ?
Nov 5 06:21:05 (none) user.warn kernel: io_callback: io failed, status = 5