[NCLUG] bootable software raid 1?

[Bob Proulx]

Erich wrote:
> I am attempting to simulate a drive failure by removing one drive,
> and replacing with a blank of the same capacity.  The cables were
> set to cable select.  I cannot seem to get the system to boot.

First some quesions.  Are you using lilo or grub and what versions.
How are you trying to boot?  Off of the raid partition or off of the
raw partition?  (Actually, this won't tell me enough information.  But
it might spark a response from another.)

In my experience using software raid on my systems I cannot boot if
the primary boot disk has been failed.  I can boot if only the
secondary disk has been failed.  If I swap the good secondary into the
failed primary position as the primary boot device then I can boot
fine.  This is a limitation of the version of lilo that I am currently
running.  I cannot boot of off the raid partition.  I can only boot
off of the raw partition.  Which means I have to pick one and I don't
get a raid for booting, only for running.  My raid machines are older
and running great on their older software so I have not updated for
some time.  It ain't broke so I am not going to fix it.  Your
experience with newer software may show it works better now.

I have heard that there are patches to lilo or perhaps newer versions
to enable booting off of the raid device and therefore enable booting
even if the primary boot device was failed.  (Actually in that case of
using the raid device itself to boot the whole idea of a primary or
secondary drive rather becomes moot.)  So perhaps someone on the list
will volunteer information about the history of that particular
feature and we can both learn about it.

Grub is generally considered superior to lilo in many ways.  But I
have not converted due to laziness and the fact that what I am running
is working good enough.  There are just too many things to do to get
them all done.

I software raid all partitions including swap.  This way a failed disk
will not crash a machine.  Since I don't automatically reboot this is
enough for me.  I should not ever see a machine crash due to a failed
disk.  When a disk fails I get an error message by email and I can
address it on my own schedule.  If I am rebooting a machine and it
does not boot due to a failed primary I can always swap drives
manually and get the system up.

Bob