Dongsheng,
I've been able to reproduce (what appears to be) your problem on my i386
kvm box.
In my case, using CPU scaling (apmd) on my OpenBSD client would cause it
to get screwy and lose the clock source if I alternately let it run idle
and then ran a bunch of CPU or I/O intensive tasks on it.
Some googling appears that this is a known problem with some versions of
KVM not emulating the RTC correctly, though I'm not 100% sure, as I'm
not a KVM developer. Folks have recommended disabling HPET in the BIOS
and/or making sure that the host linux kernel has HPET_EMULATE_RTC set
in the .config though I experienced this problem with my OpenBSD guests
even with that option set in the hosting linux kernel.
In my case, I found that not enabling apmd at all (and rebooting the
guests where I *had* enabled it, even once) made the problem much more
bearable. Basically if the machine is mostly idle I never get the
"Alternate system clock has died" message on 'systat -w1 vmstat', and it
doesn't lock up anymore like it did when apmd was running.
Now the only issues I have are that under high load I occasionally get
em0: watchdog timeout -- resetting
em0: watchdog timeout -- resetting
em0: watchdog timeout -- resetting
messages on the console when using the e1000 NIC emulation, but it
doesn't seem to be too severe right now, and is only occurring when I'm
serving a decent bit of data (55-68mbps SSH/rsync data).
Best of luck,
-Tico
Dongsheng Song wrote:
When I running OpenBSD under kvm, process time aways 0 !
[dongsh...@dl:~/kvm]% cat OpenBSD-x64/start.sh [09-01-05 21:53:50]
#!/bin/sh
cd /home/dongsheng/kvm/OpenBSD-x64
kvm -name OpenBSD-x64 -m 1024M -hda hda.img \
-cdrom ../../var/iso/openbsd-amd64-4_4-20081215.iso \
-net nic,vlan=0,macaddr=52:54:00:12:34:01,model=e1000 \
-net tap,vlan=0,ifname=tap01,script=no \
-net nic,vlan=1,macaddr=52:54:00:12:34:11,model=e1000 \
-net tap,vlan=1,ifname=tap11,script=no \
-vnc :11 -daemonize
[dongsh...@x64:~]% w [09-01-05 21:53:17]
9:53PM up 16 days, 13:30, 1 user, load averages: 0.08, 0.08, 0.08
USER TTY FROM LOGIN@ IDLE WHAT
dongsheng p0 116.23.101.68 9:53PM 0 w
load averages: 0.06, 0.08, 0.08 21:53:37
17 processes: 16 idle, 1 on processor
CPU states: 1.1% user, 0.3% nice, 11.0% system, 7.9% interrupt, 79.7% idle
Memory: Real: 11M/134M act/tot Free: 852M Swap: 0K/2055M used/tot
PID USERNAME PRI NICE SIZE RES STATE WAIT TIME CPU COMMAND
8430 root 2 0 1188K 2032K sleep select 0:00 0.00% sendmail
26190 root 2 0 696K 1344K idle select 0:00 0.00% sshd
26396 _syslogd 2 0 452K 824K sleep poll 0:00 0.00% syslogd
26716 root 2 0 472K 884K idle select 0:00 0.00% cron
16624 root 2 0 400K 868K idle select 0:00 0.00% inetd
1 root 10 0 360K 364K idle wait 0:00 0.00% init
21396 dongshen 2 0 340K 1416K idle select 0:00 0.00% ssh-agent
13401 dongshen 18 0 976K 3132K sleep pause 0:00 0.00% zsh
10013 root 2 0 3372K 3092K idle netio 0:00 0.00% sshd
7279 root 2 0 420K 740K idle netio 0:00 0.00% syslogd
20347 dongshen 28 0 448K 1496K onproc - 0:00 0.00% top
15854 dongshen 2 0 3344K 2180K sleep select 0:00 0.00% sshd
2009/1/4 Michiel van Baak <mich...@vanbaak.info>:
On 09:41, Sat 03 Jan 09, Daniel A. Ramaley wrote:
Running OpenBSD under VirtualBox is not stable at all.
I have good experience running OpenBSD under xen, kvm and vmware-server.
