ML110G6 extend raid on b110i

[mastsrl]

i need to setup a ML110G6 with 2 disk initially that'll later will need to be expanded to 4.
i've done lab tests and the ACU does not have any option to extend the raid(even when it's set as raid10 initially).

also, if i do create two raid 1 arrays and then do a mirror on software(windows 2003) then i end up with a raid 1+1+1+1 which equals to the size of ONE HDD

any ideas?

[mastsrl]

one idea i have is to create the raid 10 from the go using 2 loaned HDDs, then remove those drives and replace with the definite ones, but i'm unsure how will the b110i react to such a change(i'll have to scrounge some HDDs to test... 6 of them)

[TTr]

In ACU, the option to extend the raid will not be available until you have connected the two extra drives. However I don't know if the B110i has that feature, I could not find it in the manual or the server quickspecs.

http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c01706551/c01706551.pdf

http://h18004.www1.hp.com/products/quickspecs/13477_na/13477_na.pdf

You could create two separate raid1 arrays (mirrored pairs) on the B110 and do software striping in windows (if capable) or vice versa two raid0 arrays on the B110i and mirror them in windows. This way you will end up with the equivalent space of two disks.

Doing the "cascaded mirroring" as you are suggesting will give you the space of one disk and it is an overkill.

[mastsrl]

some more testing done:
it appears the b110i does not support any kind of capacity expansion(either on/offline).
i tested this by adding two more blank disks and the only option was the create a new array.

Since windows 2003 only supports adding a mirror(i could create a stripe, but for converting the system volume, the only option is mirror) i could mirror two raid 0 arrays but i'm not particularly keen on that idea as it has too much risk of loosing 2 drives.

Also are we sure that if one drive fails on a raid0 array the controller will block access fast enough to prevent upstream corruption?, i mean, windows only sees two volumes, if one starts putting out garbage before the controller drops array access(IF it does, i haven't tested raid0 failures in this manner) and windows fails to notices it and synchronizes across the array then i'll have a huge mess.

Right now i'm testing the following scenario:
RAID 10 4xHDD created --> 2 interim HDD removed -->2 blank definite HDD added -->rebuild

i'm at stage 2, so far removing 2 drives hasn't affected the system, i'm about to add 2 replacements and see what happens...

[mastsrl]

definitively works, adding two "replacement" HDDs triggered a complete rebuild, with high priority rebuild it even took ~15min to finish(clean win2003 install)

granted i only removed sequential drives(ports 3 and 4 removed), i didn't had the time to test mixed removal(like 1&3/2&4) but it will essentially be the same due to RAID 10 function so it's safe to say it's inconsequential.

btw, did some fast benchmarks, with 2x160GB(different families even) one 250GB and one 500GB(with remapped sectors) and the average was 132MB/s, something was limiting sequential as it ran "flat" from end-to-end with several downsipkjes and oddities but it's the system drive so i'm not surprised much

[mastsrl]

b110i cannot extend raid online or offline.
solution: build raid10 with borrowed HDDs, then replace the 2 borrowed HDDs with the definite ones and rebuild