[OpenIndiana-discuss] iostat

[jason matthews]

i am starting to wonder if iostat is broken in 151a8 - perhaps for this class of disk? in this pass i am testing a single mirror. the  filebench parameters are set for 8k random writes.

the numbers reported by iostat differ wildly from filebench, zpool iostat, and fsstat zfs which seem to express similar values over the same time period.

iostat returns 11541 writes/s @ 314 MB/s which would imply that the write size was 39250 bytes (314MB/8k)
zpool iostat consistently returns about 22-23k write ops/s at about 163MB/s
fsstat zfs returns about 20k write ops/s at about 163MB/s
filebench returned for the whole run 

any thoughts?

j.