As BIll said, iowait as reported by mpstat et al is zero on Solaris 10; iowait
is only a useful metric on a uniprocessor machine, its value is not useful when
there is more than one processor in the system. Since it's been the center of
much confusion, it's been removed.
A better way would be to measure the time spent waiting of the thread of
interest; i.e. look at the amount of time spent waiting for I/O by the "find"
command. You can do this with DTrace quite easily:
sol10$ ./iowait.d 639 ^C
Time breakdown (milliseconds):
<on cpu> 2478
<I/O wait> 6326
I/O wait breakdown (milliseconds):
file1 236
file2 241
file4 244
file3 264
file5 277
file7 330 . . .
Regards,
Richard.
#!/usr/sbin/dtrace -s
# iotime.d
#
#pragma D option quiet
BEGIN
{
printf("%10s %58s %2s %8s %10s\n", "DEVICE", "FILE", "RW", "Size",
"Time");
last = timestamp;
}
io:::start
{
self->blkno = args[0]->b_blkno;
self->dev = args[0]->b_edev;
self->flags = args[0]->b_flags;
blockio_start[self->dev, self->blkno] = timestamp;
printf("%10s %58s %2s %8d %-10d\n",
args[1]->dev_statname,
args[2]->fi_pathname, args[0]->b_flags & B_READ ? "R" : "W",
args[0]->b_bcount,
(uint64_t)(timestamp - blockio_start[self->dev,
self->blkno]) / 1000);
last = timestamp;
}
io:::done
/ blockio_start[args[0]->b_edev,
args[0]->b_blkno]/
{
self->blkno = args[0]->b_blkno;
self->dev = args[0]->b_edev;
printf("%10s %58s %2s %8d %-10d\n",
args[1]->dev_statname,
args[2]->fi_pathname, args[0]->b_flags & B_READ ? "R" : "W",
args[0]->b_bcount,
(uint64_t)(timestamp - blockio_start[self->dev,
self->blkno]) / 1000);
@a[self->dev] = lquantize((timestamp -
blockio_start[self->dev, self->blkno]) / 1000,
1000, 500000, 1000);
}
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
