Time is critical.

In VMs this criticality is even more pronounced. Time slips… CPU instructions go askew, and things get weird. That said, there are situations when you may wish to disable the built in VMware Tools Time sync service… completely. What do I mean by completely? Well, even with the tools time sync set to disabled, VMware will force a VM to sync with the host in some specific situations, and despite my valiant Google effort, I was unable to uncover exactly what these are.

According to KB 1189, you can disable TimeSync completely by adding a few parameters to the config file:

If you want to completely disable time synchronization in the guest, open the virtual machine’s configuration file (.vmx) in a text editor and set the following options to zero, as shown below.

tools.syncTime = "0"
time.synchronize.continue = "0"
time.synchronize.restore = "0"
time.synchronize.resume.disk = "0"
time.synchronize.shrink = "0"
time.synchronize.tools.startup = "0"

While that is borderline awesome… when combined with Glenn’s script from get-admin:

$ExtraOptions = @{
    "tools.syncTime"="0";
    "time.synchronize.continue"="0";
    "time.synchronize.restore"="0";
    "time.synchronize.resume.disk"="0";
    "time.synchronize.shrink"="0";
    "time.synchronize.tools.startup"="0";
}   # build our configspec using the hashtable from above.  I prefer this
# method over the use of files b/c it has one less needless dependency.
$vmConfigSpec = New-Object VMware.Vim.VirtualMachineConfigSpec
# note we have to call the GetEnumerator before we can iterate through
Foreach ($Option in $ExtraOptions.GetEnumerator()) {
    $OptionValue = New-Object VMware.Vim.optionvalue
    $OptionValue.Key = $Option.Key
    $OptionValue.Value = $Option.Value
    $vmConfigSpec.extraconfig += $OptionValue
}
# Get all vm's not including templates
$VMs = Get-View -ViewType VirtualMachine -Property Name -Filter @{"Config.Template"="false"}   # Do it!
foreach($vm in $vms){
    $vm.ReconfigVM_Task($vmConfigSpec)
}

There you have it. Time disabled in all your VMs.

What is a “Step Ticker”

You know, it’s one of those little pedometer things you strap onto your hip and tracks how far you’ve walked. Ok, so it’s not that. At least not within the context of servers & virtualization. So what IS it? It’s a horribly named NTP (network time protocol) concept. Yup. That’s it. You were expecting at least two other paragraphs, no?

What does it do?

Hate to disappoint, this paragraph is going to be short too. A Step Ticker is basically a way to tell the NTP daemon to check a server or two, and correct time on startup, rather than relying on the time that was restored from the hardware.

This is meaningful, because often the hardware time has drifted. If you are running applications that are exceedingly time sensitive, like virtualization, you’ll of course want accurate time. So setting up Step Tickers is a “Good Idea”. So how do we do that?

Setting Up Step Tickers

I’ll show you how to do this on a single host, as well, that’s easy, and I’m lazy. To deploy in mass, you’ll likely want this as part of your kickstart process. If you don’t have automated builds, shame on you, but you should have it as part of your QC process then. As for scripting the change out to each host, that is best left to you, the reader (but can be done in PoSH with plink).

Enough rambling, open /etc/ntp/step-tickers and add some lines that look like this

0.vmware.pool.ntp.org
1.vmware.pool.ntp.org
2.vmware.pool.ntp.org

From here, save the file and restart ntp (service ntpd restart).

Questions? Drop a line in the comments.

Sources: Tripwire KB, VMware KB, and some random UseNet/Mailing list  post about ntpd.

VMware has updated their Time Keeping Best Practices KB’s.

The new timekeeping articles are:

1006427

Timekeeping best practices for Linux

1007020

Linux using TSC clocksource stops responding

1008284

Time falls behind in a virtual machine when the guest operating system writes to previously unwritten regions of its virtual disk

1005861

Time falls behind in virtual machine when the memory of the virtual machine paged from disk by the VMkernel

1006072

Time in a virtual machine drifts due to hardware timer drift

1006113

Time in a Linux 2.6 guest operating system runs faster than real time due to lost tick overcompensation

1005953

Time runs too fast in a Windows virtual machine when the Multimedia Timer interface is used

1006090

Time drifts in the virtual machine and the service console due to the HPET misreporting its frequency

1006086

Time in a Linux virtual machine jumps backward when using clock=pit

1006088

Time runs slower than real time due to lost timer interrupts

1005802

Determining and changing the rate of timer interrupts a guest operating system requests

Had a situation in which time in a Linux guest kept creeping about, slower, faster, etc. To the point where NTP wasn’t helping before. This in turn, like all good questions, turned me to that which is the holy oracle of all knowledge: Google.

Google, showed me that I am not alone in my struggles, and gave me some direction in my quest. That direction was to VMware’s KB 1006427: Timekeeping best practices for Linux

It seems as each distribution does things slightly differently, there are kernel parameters that can be set to overcome the time slippage.

This article presents best practices for Linux timekeeping. These recommendations include specifics on the particular kernel command line options to use for the Linux operating system of interest. There is also a description of the recommended settings and usage for NTP time sync, configuration of VMware Tools time synchronization, and Virtual Hardware Clock configuration, to achieve best timekeeping results.