Clock on CentOS

Update, August 2010

Note that allegedly this is all more or less depreciated. Kernels later than those provided in 5.2 have more vm-aware stuff in them; therefore there's less a requirement to mess around with this. I have found that CentOS 5 VMs based on 5.4 or later only lose time if the VM host is overloaded somehow. For "routine" virtualization, there shouldn't be any problems running NTP with a stock installation.

VMware Solution

See this page for VMware's take on how to deal with this. This appears to be the preferred solution, rather than what is below.

Problem

I have VMware Server 1.0.6 installed on a Windows 2003 server. I have CentOS 5.1 installed into a VM. The clock on my CentOS 5 system is incredibly bad (ie loses 30 seconds per minute bad).

Most recent recipe

1. Disable cpuspeed and friends on host OS (for Linux hosts)
2. Kernel from http://vmware.xaox.net/centos/ (alternatively: try the boot kernel option divider=10 for CentOS 5 guests (Warning: I hosed at least one VM by using this kernel options line. The VM would answer pings but ran so slowly and erratically that it would not hold a network connection so we couldn't SSH into it to fix it))
3. See Installing vmware-tools on CentOS
4. Boot with kernel options:

   ro quiet clocksource=pit nosmp noapic nolapic

Commentary

See the various VMware links I've been accumulating at http://del.icio.us/xdroop/vmware

Alleged Solution

1. See Installing vmware-tools on CentOS
2. Add

   clock=pit

   to the kernel line in your grub.conf
3. Now on your HOST, edit your guest's .vmx file and change the line from:

   tools.syncTime = "FALSE"

   to "TRUE". In mine, it's the last line in the .vmx file.

   # vmware-cmd /path/to/machine.vmx getconfig tools.syncTime

4. Restart the VM.

(Ref)

Update: this doesn't work. The kernel magic is

clocksource=pit