> Hmm...the iostat data does not make sense. > Consistent non-zero values (12/13) in the wait > queue, with an IO rate of zero (no read, no writes).
This can occur if an I/O is taking a very long time to complete, or gets lost. Since it doesn't complete, we can't count it as completed, so it drives the utilization metric up. For typical systems, you shouldn't see this, so I would check the hardware for faults -- I would expect to see some timeouts/retries. -- richard -- This message posted from opensolaris.org _______________________________________________ dtrace-discuss mailing list dtrace-discuss@opensolaris.org
