Time Drift on ESXi When Using Windows Server as NTP Server

[Fujitsu_Admin_D]

Hello

We use a Standalone Windows Server (2019) as the time source for our environment.  It is configured as an NTP server.

Windows DCs, Network equipment and VMware (ESXi) are configured to use this as the NTP time source.

All work OK, but we are noticing some drifting with the ESX servers.

After some time, we notice that the ESXi servers are not in sync with the Windows NTP Server.  If we stop and start ESXi NTPd, the time syncs as expected.

However, after some time passes, we notice the time drifts on the ESXi.

What is causing ESXi not to remain in sync?  Apart from restarting NTPd, is there any solution to this?

Kind Regards

[jlanders · Accepted solution]

There are two popular options. First, you can stay with Windows Server as your upstream time provider. You'll want to look at this KB article (https://kb.vmware.com/s/article/1035833) for hints on updating ESXi's NTP configuration to deal with Windows Server's looser time accuracy as an NTP source.  See the part about adding a "tos maxdist 30" line to your ESXi's NTP configuration in VMware's KB article. In general, ESXi's open source NTP client may reject time packets if the upstream time source signals poor accuracy. Adding the "tos maxdist 30" line allows ESXi's  NTP daemon to accept packets over a much large accuracy range.

Second, you can select a better, compatible NTP time source and point both your ESXi hosts and your Windows AD Server at that time source. Microsoft says the latest versions of Windows Server can be configured for better accuracy. Look for their article titled "Configuring Systems for High Accuracy" in the documentation on their site. It includes information on configuring the Windows time service that may be helpful for this option.