Hi all I have just got my hands of two servers with 2TB of internal storage (Petabox GB2000) - these servers are going to be used for storage in a way were the files placed on one of the server is going to rsynced to the other. Given that I then achieve redundancy on server level and that each server contains 4 disks, I am in the mood for deploying software stripe across these 4 disks for maximum throughput.
Before doing so I am however a little unsure of the stability of software stripe, I am of cause aware of the fact that loosing a disk will mean loosing the entire stripe set but given I have a replicata this is not my main specultaion, what I want to know is if software stripe is stable or if it have some insane problems (I remember in kernel 2.4 days where a power failure could take down a software mirror set - so if this can happen to a RAID 1 then I am a bit afraid of what might happen to a RAID 0). Another thought is on the ease of rescuing a OS on a software stripe (i.e. if I somehow mess up my kernel or some other key software) - on a normal hardware raid/disk I can just boot using Knoppix and fix whatever problems I might have created - is such operations extremely difficult/impossible on a software stripe ?. Another solution would be to place the OS on 1 disk and then stripe the remaining 3 disks and link the remaining storage on disk 1 together with the stripe on disk 2, 3, and 4 using LVM - this way I can boot into the OS using normal forensic tools and still achieve greater throughput than just chaining the 4 disks together using LVM (as in without the software stripe). Regards. Lars Roland
