Information about creating the Emergency Boot Diskettes :

• In UnixWare7, you can't create "generic" boot/root floppies for emergency recovery. In fact, they are not even boot/root floppies anymore. They are both the boot floppies.

  Things are handled a little bit differently. Use this command to create them :

  • "/sbin/emergency_disk diskette1"

  You may however run into a problem making these diskettes. I have worked on many of these machines and most had no problems. My personal Asus SCSI 7980 machine was no problem, our office Acer 9100 Altos was no problem, a Dell 2300 non-raid was no problem, and a Gateway/ALR 7890 was no problem. However we had problems on an IBM PC Server 330 model, with IBM's IPS raid installed. Seems we were getting errors while trying to create the disks. I have discovered from the error logs created, that we need to create a link, for a non-existent directory. Do this by :

• cd /usr/lib/drv/locale
• ln -s C ./en_US

Remember that the creation of the emergency disks, must be done in single user mode !!.

How do you recover from a fatal system crash without the boot floppies you ask?

Well since we already ran into this, and NOBODY seems to know how, please read on :

• The only way to get around this seems to be to begin another installation and break out to a shell. I had tried making emergency disks from a 7.0.1 SCSI 7880 machine, but it would not allow access to my disks on the IBM PC Server with IPS raid installed. I know it was due to the proper hardware drivers not being loaded, but I was in a quandry on how get those loaded. My emergency disks from a version 7.0.0 SCSI 7880 machine, would not even prompt for the second disk !! I had tried all kinds of things, because I couldnt use the ls and find commands on the partial boot from the wrong emergency disks, so I cpio'd these files off an existing 7.0.1 installation, and then could use them. This was all just to find out that I couldnt access the disk anyway. Forget mounting, the machine didnt even know about them. This makes sense due to the hardware driver problem.

  So how do you load them you ask? Well after searching my favorite trouble shooting tool, dejanews, for about an hour, I found that many, many people had similar questions, but NOBODY had an answer. Everyone seems to reload the O/S. This does NOT sound like an option to me. In my case, it was a brand new install, so I could have, but what the hell do you do if it is an existing single disk installation, with much data? Well here you go, and dont worry, this wont affect any information stored on the disks unless you do it explicitly after mounting. The install process never updates anything, we just use it to load our drivers temporarily :

  1. Insert the installation disk, disk 1 of 2, and power on the machine.
  2. Insert the second installation disk, disk 2 of 2, when prompted.
  3. Press F8 to defer licensing for now, and confirm that we will use an evaluation license.
  4. Insert the HBA disk when prompted. I had to do the IBM IPS Raid diskette first, then the SCO UnixWare7 HBA disk second, since my CDRom was on an Adaptec controller, and we will need access to our CDRom drive.
  5. Once you have finished all your HBA loads, make sure to skip the DCU.
  6. When you are prompted for the node name, hit <CTRL><ALT><ESC> to break into a virtual terminal.
  7. I am assuming most of you took the vxfs file system format for its resiliency, and hence our problem. What we need to do now, is fsck the root filesystem, because I would strongly suspect, your machine didnt go down nicely either. But to fsck the file system, we need to have the vxfs kernel module loaded... but to load the kernel module, we must first register it.... but to register it, we will need the cdrom. So insert the installation cdrom disk 1 of 3 into the drive.
  8. Mount the cdrom by typing : mount -Fcdfs -r /dev/cdrom1 /cd-rom , note that the /cd-rom mount point is already provided.
  9. Now we can register the vxfs kernel module by typing : modreg 4 vxfs , note that modreg is a shell function not a command, and there is no damn help on this command anywhere that I can find. If someone else finds help for it, let me know.
  10. Well, now that its registered, let load it, by typing : modadmin -l /cd-rom/.extra.d/etc/conf/mod.d/vxfs , and note that you need to change cd-rom for whatever mount point you used, if you didnt follow my "expert" advice.
  11. Now that the kernel drivers are loaded, we can fix the damn disk. Finally, type : /cd-rom/.extra.d/etc/fs/vxfs/fsck /dev/dsk/c0b0t0d0s1 , noting that the root file system, like in NCR/AT&T speak for all of you guys with that base, is always slice 1.
  12. Now that the file system is fsck'd, you can mount it by typing : mount -Fvxfs /dev/dsk/c0b0t0d0s1 /mnt , noting that the /mnt mount point is already provided for you.
  13. A nice feature of following these mount points, is that now that everything is mounted, the /mnt prefix is already in your path, so most of your real root file system command can now be used. Personally, I needed to edit a buggy /etc/rc2.d script, that was hanging the machine up on startup, but your problem may be different.
  14. Now that you have (hopefully) fixed your problem, you need to just reboot the system, and all is well. Dont forget to remove the floppies, and CDRom's from their appropriate drives.

More info to follow on more problems ....

This page last updated on 04-19-1999.