I hope not to bother beyond the limit, but the security of mirror raid is something of utmost importance, at least in my work of biochemist, with very limited ability in recovering from disk failures.
I planned to use the double-opteron, two sockets, server, tya 64, as a victim for the test you suggested. However, the test
root@tya64:/home/francesco# dd bs=512 count=1 if=/dev/sda 2>/dev/null | strings
ZRr=
`|f
\|f1
GRUB
Geom
Hard Disk
Read
Error
root@tya64:/home/francesco# dd bs=512 count=1 if=/dev/sdb 2>/dev/null | strings
ZRr=
`|f
\|f1
GRUB
Geom
Hard Disk
Read
Error
suggested that grub was installed on both disks. Using one disk at a time, as you suggested, was in accordance.
root@gig64:/home/francesco# dd bs=512 count=1 if=/dev/sda 2>/dev/null | strings
ZRr=
`|f
\|f1
GRUB
Geom
Hard Disk
Read
Error
root@gig64:/home/francesco# dd bs=512 count=1 if=/dev/sdb 2>/dev/null | strings
root@gig64:/home/francesco#
indicating that grub is installed on sda only. Confirmed by using one disk at a time.
500GB disks on tya64 came from a dismissed doubleOpteron four socket server that I had assembled several years ago. Probably at that time in my activity as a biochemist I was left more time to be careful about linux. With gig64 the two 1000GB disks came recently from the store and the described failure as to installing grub on sdb was with them. Unless the problem is different, related to the particular gig64 machine.
QUESTIONS:
(1) If there is any chance that the particular hardware of gig64, comprising two GPUs (I called gig64 a server, while, unlike the real server tya64, it is a consumer mainboard GA-X79-UD3 with Intel(R) Core(TM) i7-3930K CPU @ 3.20GHz and two GTX680 GPUs, leading to very fast number crunching, NOT overclocked) can interfere with installation of grub on the second disk, it seems to me that gig64 is the appropriate victim. However, by replacing the two 1000GB disks with two spare victim disks that I should have somewhere (amd64 on both, and likely grub too). If anything, linux is unable to tell anything about the GPUs (and even nvidia-smi tools tell very little, which my lend suspicion on why the grub installation of the second disk failed). Fundamentally, it is a game machine, so that no chance to get mirror raid even mentioned by bloggers of this type of computers.
(2) Is the above pipe test (that grub installed leads to some message when failure is encountered, while no message means no grub available) always reliable and equivalent to detaching disks?
thanks
francesco