Timothy Coalson <[email protected]> writes: > On Fri, Mar 24, 2017 at 5:43 PM, Harry Putnam <[email protected]> wrote: >> >> On the vm, I've created discs like so: >> >> 2 @ 96G >> 2 @ 116G >> 2 @ 216G >> >> These have just over 2x variation in size, and you put them all in a > single raidz vdev, that is why your space is "missing" - it is just like > making a mirror out of a 96G disk and a 216G disk, which would have only > 96G available and "lose" over 100G of space. raidz doesn't do anything > magical for different-sized disks - it uses a slice from each device that > is equal to the *smallest* device in the vdev. > > So, your raw space *before* parity or overhead (which is what "zpool list" > says) should be 6*96G = 576G. Remove 1 device worth for raidz1 (5*96G = > 480G), and a fudge factor for metadata overhead, and lands you at the 460G > that "zfs list" says about it.
Wow, I really have missed the boat here. I did not realize that raidz would reduce everthing to lowest common denominator. I won't be so `tricky' with disc sizing from here on. The sad part is since it is vm stuff, I could of just as well used all 200G dics and had 1200 g raw space ... or even better... 2 600g discs in mirrored arrangement, and forget all about raidz. _______________________________________________ openindiana-discuss mailing list [email protected] https://openindiana.org/mailman/listinfo/openindiana-discuss
