SCSI and IDE on RH 7.1

Robert Canary phantom at ohiocounty.net
Mon May 23 03:15:57 UTC 2005


I had a SCSI system to fail this week on one of our low end servers.  This was a 
SuperMicro 1U with an Adaptec 7899 builtin on the motherboard.

Both drives have CRC and bus reset errors, however, after some long hours of research I 
found the Seagate Cheetah drives have a bug in the firmware and they needed to 
upgraded.(hopefully the data on the drive is still okay)

Meanwhile I installed a 129G EIDE drive and reloaded the RH7.1 Boxset.  The system booted 
fine but I still couldn't get to the SCSI drives.

Amazingly the up2date still worked and all my packages were updated accordingly, however, 
I also updated the kernel.

Now here comes the quirky thing.....definetly a bug

I can boot on the original kernel(2.4.2-2), but if I bootup on the up2date 
kernel(2.4.20-28.7) then it will always fail on boot up with filesystem error on "/" and 
drops me to a "repair filesytem" prompt.  However, if I disable the SCSI then it will boot 
fine.  No matter how many times I fsck the "/" it will continuely fail.

Now it really gets wierd....
Although the system is boot from /dev/hda and the boot partition is on hda1 it is scanning 
/dev/sda1 as the root file system.   /hda1 has no reference to /sda1.  And of course the 
scan will always fail because of a failed SCSI system.  If I remove the sda and sdb drives 
then it will boot fine (these load from the front).

What is causeing the 2.4.20 kernel to scan the failed sda instead of the correct boot 
partition of hda1?

I also tried disabling the onboard Adaptec and installed an adaptec 29160 controler (which 
loads a different driver) on the PCI bus, but it dose the same exact thing.  So it is 
*not* the Adaptec Driver or the Adaptec hardware causeing the problem.

--
robert




More information about the redhat-list mailing list