Bart Smaalders's email at 5/30/2006 10:04 PM, said:
Bill Rushmore wrote:
I am running build SXCR 39 on my Sun Blade 100 and the performance on
it is horrible lately. How horrible you ask? JDS is almost unsable
because the mouse is jumpy and typing lags. The uptime load average
is over 4 at idle for JDS. But how about this as a benchmark for
building build 40:
==== Total build time ====
real 44:57:10
The longest this has taken a few builds back was around 10 hours.
Nothing else of any significance was running when I did the build. I
did notice using top that during the build the CPU was at 100%, as to
be expexted, most of the CPU was being used by the kernel, usually
over 60% and around 30% for user processes and load was pretty much
over 6 the whole time. One difference I have with build 39 on this
box is that I am running ZFS for my home directory (compression is
turned off). Could ZFS cause that much overhead? I am doubtful so I
don't want to go fdisking my machine just yet. I would appreciate any
suggestions for troubleshooting the issue. With DTrace it should be
easy to discover the problem but I need some advice on where (and
maybe in how) to start.
This message posted from opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
I think this is a known bug in the sparc ATA driver; are you
building on an IDE drive?
6421427 netra x1 slagged by NFS over ZFS leading to long spins in the
ATA driver code
If you're spending a lot of time in ata_wait, this is the
problem.
=- Bart
When you are running high %sys, run this (^C after about 10 seconds) and
send the output:
# dtrace -n profile-501'[EMAIL PROTECTED]()] = count()}END{trunc(@, 25)}'
Thanks,
Jarod
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
