Linux and software-raid are new to me so I thought I'd run my plans by
experienced users to help me find holes in my understanding and thinking.

I am building a web server using RH 6.0 boxed on a system with 3 IDE drives.
I want to use software raid-1 mirroring on two of the drives. I plan to use
the third drive for a combination of a configuration/emergency system and
to provide a mountable filesystem for storing system and application
archives for disaster recovery.

I hope to use raid-1 on the prod system for all but the /boot and swap
filesystems.

It should look as follows:

pri-mast -> hda:  prod sys: 
  -/boot, swap1 
  -plus "left mirror": /, /usr, /home, /var, /var/lib/mysql
pri-slave-> hdb: config/emergency sys
  -plus /backup (tar archive) filesystem
sec-mast-> hdc:  prod sys (right mirror)
  -/boot, swap2
  -plus "right mirror": /, /usr, /home, /var, /var/lib/mysql


Recovery scenarios I've thought of are as follows:

o If I loose one of the raid drives and my system survives (unlikely due to
swap):
-continue to run on surviving drive until maint convenient. Then:
-Power down
-replace failed drive
-boot emergency system (hdb using Lilo if hdc failed -or- loadlin.exe if
hda failed "loadlin vmlinuz root=/dev/hdb8")
-mount and tar backup surviving file systems
-partition new drive
-rebuild mirrors
-tar restore surviving filesystems
-reboot production system.

o If I loose one of the raid drives and the system crashes (more likely)
- boot the emergency system (as above)
-edit the /etc/fstab on the surviving drive point to surviving drive
filesystems (i.e. 'md0'->'hdc5').
-boot production system (hda or hdc) using Lilo if hdc failed -or-
loadlin.exe if hda failed "loadlin vmlinuz root=/dev/hdc8")
-run unmirrored until maint is convenient
-perform 1st recovery procedure above.

o If data corruption has occurred
-boot the emergency system
-perform any necessary prepatory steps (i.e. break mirrors if necessary (as
above))
-tar restore system and application archives from periodic backup (daily)
-reboot production system
-run until maint is convenient
-perform 1st recovery procedure above (if appropriate).

What do you think?

Also, I'm having some trouble starting Linux with loadlin which I will run
from an emergency diskette which I created as follows:

from W98-DOS window -> format a: /s
from Linux ->:
mount  /mnt/cdrom
mount -t vfat /dev/fd0 /mnt/floppy
cp /mnt/cdrom/dosutils/loadlin.exe /mnt/floppy
cp /boot/vmlinuz /mnt/floppy
umount /mnt/floppy

To boot emergency linux system without Lilo:

-boot emergency diskette
-enter following command-> a:loadlin vmlinuz root=/dev/hdb8

When I enter the loadlin command I get a message saying that vmlinuz is not
an image file. This procedure came from the RH Linux Unleashed book.

Thanks for any help,
Joel Fowler

Reply via email to