Wildly out of control clock under vmware

grey0x2a · Feb 23, 2009, 6:49 PM

I am using pfsense under vmware. The clock is running very fast 1000s of seconds per hour and the ntp deamon does not
seam to correct this at all.

Any clues on how to fix this?

Thanks.
.
–
Chris

cmb · Feb 24, 2009, 1:12 AM

Are you using the VMware appliance, or have the Open-VM-Tools package installed?

grey0x2a · Feb 24, 2009, 1:34 AM

Yes I am using the vmware tools version 102166_7_1. pfsense version 1.2.2

I just check my .vmx configuration it has

tools.syncTime = "FALSE"

should that be "TRUE"? I am running it in production so I don't want an unnecessary reboot.

–
Chris

cmb · Feb 24, 2009, 1:42 AM

yeah set that to true

grey0x2a · Feb 24, 2009, 6:52 AM

No joy. Setting that parameter to TRUE did not give an noticeable improvement in clock accuracy.

Cry Havok · Feb 24, 2009, 11:06 AM

Disable any CPU throttling on your underlying platform.

grey0x2a · Feb 24, 2009, 2:47 PM

How? Where?

Is this a config option of vmware? or am I nice-ing vmware with out knowing it?

Cry Havok · Feb 24, 2009, 3:23 PM

It's a setting either in the BIOS and/or in your operating system, I don't know if VMWare server can change this.

grey0x2a · Feb 24, 2009, 5:12 PM

I checked the proccess table I am running this is what I got:
command : ps -A -o pid,user,time,nice,args | grep vmware
3192 root 00:00:00 0 /usr/bin/vmnet-natd -d /var/run/vmnet-natd-4.pid -m /var/run/vmnet-natd-4.mac -c /etc/vmware/vmnet4/nat/nat.conf
3201 root 00:00:00 0 /usr/bin/vmnet-natd -d /var/run/vmnet-natd-8.pid -m /var/run/vmnet-natd-8.mac -c /etc/vmware/vmnet8/nat/nat.conf
3207 root 00:05:38 0 /usr/sbin/vmware-serverd -s -d
3245 root 00:00:02 0 /usr/lib/vmware-mui/apache/bin/httpd.vmware -DSSL -DSSL_ONLY -DGSX -d /usr/lib/vmware-mui/apache
3251 wwwrun 00:01:34 0 /usr/lib/vmware-mui/apache/bin/httpd.vmware -DSSL -DSSL_ONLY -DGSX -d /usr/lib/vmware-mui/apache
3256 wwwrun 00:00:00 0 /usr/lib/vmware-mui/apache/bin/httpd.vmware -DSSL -DSSL_ONLY -DGSX -d /usr/lib/vmware-mui/apache
3394 root 00:00:00 0 /usr/bin/vmnet-dhcpd -cf /etc/vmware/vmnet4/dhcpd/dhcpd.conf -lf /etc/vmware/vmnet4/dhcpd/dhcpd.leases -pf /var/run/vmnet-dhcpd-vmnet4.pid vmnet4
10825 admin 01:14:55 -10 /usr/lib/vmware/bin/vmware-vmx -C /home/admin/vmware/pfSence/pfSence.vmx -@ ""
12725 admin 00:00:00 0 grep vmware

-10 is the lowest (best) priority on the system at the moment. So I do not think I am starving the vmware proccess. Should I lower the nice-ness of vmware-serverd too?

Cry Havok · Feb 24, 2009, 6:30 PM

VMWare assumes (for the purposes of time keeping) that the processor runs at a fixed speed, that has nothing to do with nice or anything else. You need to ensure that the processor does run at that fixed speed - hence disabling processor performance management in hardware (or OS).

Intel call theirs SpeedStep, AMD call it Cool 'n' Quiet. It's also known as performance states (and a few other things).

Step one - look in your BIOS for anything called SpeedStep or Cool 'n' Quiet - disable it.

Step two - what version of Linux/BSD are you running?

grey0x2a · Feb 24, 2009, 7:13 PM

Ahh ok. So I will have to reboot to try this solution.
(to check the bios)

I am running OpenSuse 11.0 on the box in question. 2.6.25 kernel. an intel x86_64 .

I just checked the "my computer" window in KDE, it said that the cpu speed was changing. So this could be the problem.

It will be a while for me to try this as the server is in production and I do not want to take it down for 20-30+ min with out
notice.