>>107716638
>The drive unevenness here is weird because both should be able to reach >170MB/s 30% in
Actually scratch that, it seems it's just an iostat accounting/presentation issue during the switch between write and read phases of the benchmark, this was the next output, which already includes a few reads:
capacity operations bandwidth
pool alloc free read write read write
----------------------------------------- ----- ----- ----- ----- ----- -----
xxxx 1.12T 2.51T 7 172 7.40M 172M
mirror-0 1.12T 2.51T 7 172 7.40M 172M
ata-ST4000VN008-2DR166 - - 1 122 1.80M 122M
ata-ST4000VN006-3CW104 - - 5 49 5.60M 50.0M
Taken together they are roughly even, but only when writing. I don't get it.