top seems to report wrong CPU usage (HVMPV)
Hugo Silva
hugo at barafranca.com
Fri May 18 11:44:01 UTC 2012
Hello,
89 processes: 4 running, 71 sleeping, 14 waiting
CPU: 77.1% user, 0.0% nice, 21.1% system, 1.8% interrupt, 0.0% idle
Mem: 151M Active, 194M Inact, 96M Wired, 3508K Cache, 60M Buf, 41M Free
Swap: 512M Total, 596K Used, 511M Free
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU COMMAND
82983 2002 77 0 115M 103M RUN 0:02 15.97% cc1plus
13 root -8 - 0K 48K - 2:20 0.98% geom{g_down}
11 root 155 ki31 0K 16K RUN 19.7H 0.00% idle
12 root -60 - 0K 224K WAIT 1:12 0.00% intr{swi4:
clock}
When running virtualized (hvm+pv, haven't really tried just hvm) I keep
noticing that there's a huge discrepancy between the %cpu top claims a
process is using vs the total cpu% being used.
In the above example, that is the view from a FreeBSD 9.0-rel-p1 hvm+pv
amd64 jailhost (cc1plus is running in a jail); -SH was given to top.
From inside the jail, we see:
CPU: 90.8% user, 0.0% nice, 9.2% system, 0.0% interrupt, 0.0% idle
Mem: 154M Active, 187M Inact, 100M Wired, 8568K Cache, 60M Buf, 36M Free
Swap: 512M Total, 596K Used, 511M Free
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU COMMAND
83454 <cut> 78 0 118M 106M RUN 0:02 19.97% cc1plus
There is a *huge* discrepancy.
I'm not sure what could be the cause. If it helps, the dom0 is running
Xen 3.3.2 on NetBSD -current.
Is anyone else seeing similar behavior from the system?
Thanks,
Hugo
More information about the freebsd-xen
mailing list