VMs on ESXi Host Not Syncing with NTP

[Justin Ferello]

I have a single Cisco UCS C240M5 running ESXi v7.0.3 build 24411414.  I currently have (3) virtual machines running on this server.  At some point, we had to power cycle this server and ever since then, my (3) virtual machines will not sync to NTP.

One VM is Cisco Prime Collaboration Deployment and two VMs are Cisco Unified Call Manager servers, both are Publishers.

We have (6) NTP appliances that are GPS/GNSS sync'ed and are running v4.

The CIMC interface on the server syncs fine to NTP.

The ESXi host syncs fine to NTP.

The VMs however will not sync to the same NTP servers, they all just say that the "host is inaccessible or down".

I can perform a packet capture on each VM though and it shows outbound and inbound traffic from the VM to the NTP appliances.

I even deployed a new CUCM VM on this host and right out of the box, just adding in the NTP servers it will not sync either.

I opened a ticket with Cisco, but they said its an ESXi issue.  I also believe there is something going on between ESXi and the VMs that is messing with the NTP packets.

Last night I updated the UCS server firmware to the latest version and upgraded ESXi to the latest 7.0.3 release, no change.

Any ideas?

I was thinking of deleting the 

[janatlud]

Hello,

there could be multiple things. Firstly, the VMs are commonly on different network than ESXi management network, so there could be some another network device/firewall in the traffic. The NTP itself could have set some firewall setting.

Second point is, from which VMs you try to reach the NTP servers. For Windows client, you have to set properly registry setting and to have allowed local policies, the firewall rules are not enough.

For unix systems, and there is the HIGHLIGHT which I need to provide, if you tested the network accessibility via "nc -vzu NTP_IP 123", there must be some bug or something, you can try this network test to non-existent device or not used IPs and it ALWAYS return "succeeded". So the point is, not to trust the NTP network test port opening from unix systems. :-)

[hgdeloitte]

Hello.

Along with janatlud's suggestion, you may also want to check the VM's time sync settings at Edit Settings --> VM Options --> VMware Tools --> Synchronize Time with Host

I have this unchecked so VMs will obtain time from AD.  Having it checked created a lot of issues.

[Andrea Consalvi]

Hi,

I had a similar issue in the past, and in one case, it was caused by the DNS settings in the Call Manager, while another time, the problem was related to the version of the NTP server providing the service.

Which DNS servers are you using? Also, are you synchronizing the VMs with an external NTP server, or are you routing through a Domain Controller?