Will and several other people are correct.
I had forgotten that ZFS does a funky form of concatenation when you use
different size vdevs. I tend to ignore this case, because it's kinda
useless (I know, I know, there's people who use it, but, really...
Basically, it will stripe across vdevs as it can.
So, if you have a zpool like this:
ZFS will have a 3-wide stripe across the first 2GB of all devices, then
a 2-wide stripe across the next 1GB of the two larger devices, then
finally a single stripe (aka no stripe) in the 1GB left in the largest
one (in this example). So you do get the full 9GB of space.
Naturally, this produces really weird performance curves for (random)
data access. OK, maybe weird isn't the right word, but...
Java System Support
Santa Clara, CA
Timezone: US/Pacific (GMT-0800)
zfs-discuss mailing list