[clue-tech] props to whoever wrote the RAID1 md code in the kernel
Nate Duehr
nate at natetech.com
Thu Apr 16 22:23:10 MDT 2009
Jim Ockers wrote:
> So the moral of the story is, Linux software RAID-1 mirroring appears to
> be a simple copy of the blocks of data between the devices, so you can
> access your data on either drive without necessarily starting the RAID
> array. I'm sure there are some drawbacks to this design (such as long
> resync times in the absence of block checksums or something) but it's
> clear to me that there are some advantages too.
I do think that was a design requirement of it, long ago...
I've also mounted up individual disks from Linux RAID1 and been able to
"see" the data.
Now if you layer it under/over LVM or something, it gets messier.
Back in the early days, many of us discussed here and on BLUG how much
of a PITA that GRUB was being, since it couldn't deal with md devices
natively as a boot device.
You had to tell it, "try from this one first... point at first disk...
then after a timeout try from this one... point at other disk". It was
brain-dead back then.
That changed and GRUB is "md aware" as long as the appropriate Linux
RAID partition type is set.
I'm old-fashioned, I still like to keep /boot separate from / and
everything else.. and right up tight against the front of the disks
physically... more chance I can fix a booting problem that way, or
recover the data from the other partition.
I dislike that many distros went to one big fat / for their default
"single-user" setups too...
Nate
More information about the clue-tech
mailing list