Hi there, just want to share some nasty experience (and what to do to solve it). Two weeks ago I installed a Raid-5 on two IDE and one SCSI partition under SuSE 6.3 with the raid0145 patches. The SCSI drive hangs of an adaptec controller. The drives were type fd and autodetection of raid is on in the kernel. md is compiled into the kernel (not a module), although the system boots from a 'normal' ide ext2 partition. I left the machine running after fdisking the drives, making the raid and installing ext2 (would liked to have installed reiserfs but readers of this group probably know why I did not). The people the machine belongs to tested and were happy. Yesterday I was again at there office and looked at /var/log/boot.msg to find that after a reboot the raid was running in degraged mode. The reason: the adaptec code was compiled as a module (standard for SuSE), during boot the md software first detects only two raid5 partitions and starts the raid in degraded mode. Only later on it loads the SCSI module, but then the raid is already detected and running. I decided to compile the adaptec code into the kernel (not as module) but that gave big problems after rebooting, SCSI errors on (a non-existing) host #1. I removed the aic7xxx module from /lib/modules/2.2.13, rebooted but that did not help. I remembered finding a new setting INITRD_MODULES in the /etc/rc.config file and removed its reference to aic7xxx. Unfortunately there is no comment in the config file on what to do after you change it. It took another reboot to find out that just changing that and running lilo doesn't help. You have to run mk_initrd to update the /boot/initrd compressed filing system that is loaded by the kernel as the initial filing system. After that the module and the compiled in kernel would no longer compete for the one SCSI device. Unfortunately this whole stuff is not in the manual, but there are some updates ( http://sdb.suse.de/sdb/de/html/errata-63-d.html and http://sdb.suse.de/sdb/de/html/adrian_6.3_boot.html ), these are in German, but you might also find some English version. Of course I only found them after solving the problem :( The new way of booting SuSE uses makes sense, but there could be some more pointers (in /etc/rc.config ) or better some automated check to see if the /boot/initrd file doesn't clash with the kernel (or its compile settings). I don't know if it is a good idea to put that in (SuSEs version of) lilo or the kernel configuration/compilation stuff. This all would not have mattered so much if rebooting a system takes so long (especially the SCSI scanning). Hope you remember this mail if you get the strange SCSI error messages and get to a solution faster than I did. Anthon
