Trace IO waits

procfs · 08-27-2008, 10:12 PM

Hi

I am have a 4 cpu 12G HP ProLiant DL580 G3, running Redhat AS3 (U 7).

When I do a top I can see some percentage if IO waits, can some one tell me or point to a document the will tell me how to do a trace on the process or processes which causing this IO wait

by the way the CPU utilization is over 90%

Thanks

Best regards

syg00 · 08-28-2008, 12:01 AM

Just because you see "iowait" doesn't mean anybody is actually waiting on I/O. Semantics and statistics - what a mix ...
iowait says the CPU is idle, and there is I/O yet to complete. Possibly independent (and non related) events.

If you want to see what tasks are in uninterruptable sleep (usually disk I/O) try this

Code:

top -b -n 1 | awk '{if (NR <=7) print; else if ($8 == "D") {print; count++} } END {print "Total status D: "count}'

To find out what task(s) is/are doing the most I/O, you'll need a recent kernel with taskstats enabled - try iotop.py or collectl.

procfs · 08-28-2008, 02:58 AM

hi syg00

Thanks for the reply, but I am new to all this is there a document that I can refer to, at least get me started