I'm running a small raid10 mdadm array that looks like follows:
# cat /proc/mdstat
Personalities : [raid10]
md0 : active raid10 sdd[3] sdc[2] sda2[0] sdb2[1]
5859211264 blocks super 1.2 512K chunks 2 near-copies [4/4] [UUUU]
First off, I am assuming the order in which the devices appear does not mean anything, but the numbers in [] behind each device tell me the actual order in the array. Is that correct?
If this is so: does this mean that sda2 and sdb2 mirror each other, and so do sdc and sdd because they are devices number 0-1 and 2-3 in the array - meaning e.g. sda and sdc can fail at the same time and my array would keep running? Or how can I find out about which is the mirror and which is the striping part?
Also, I have two SATA controllers in my small home NAS server (one onboard, one with an extra PCIe card), each with 2 SATA ports. Right now sda2+sdb2 are on the same controller, and sdc+sdd are on the other. Should I rather have one disk of each mirror on each of the two controllers? My thinking says the latter would be better because even if one controller fails, the RAID would still keep running because the striping of the two disks. Additionally I might get miniscule performance benefits because a write or read of a specific location in the array will always be using the two controllers in parallel (and thus two PCIe x1 lanes in parallel). Is this thinking correct? Can I simply shut down the machine, swap the SATA cables, and mdadm will still find the disks and assemble the array without a hitch?
