ESXi host does not sync with NTP - causes HA config failures?

[DyJohnnY]

Hi,

We are in the process of implementing some Cisco UCS systems. We using Cisco UCS blades B200-M1 (x5670, 48GB RAM)

All seems to work fine except these few things, which I believe to be related.

1. NTP syncronization is not happenning, or happening very slowly.

2. A 10 host cluster, when configured for HA, has a lot of hosts that fail to be configured for HA (example we only had 2 hosts successfully get configured for HA out of 10).

Initially I thought item 1 and 2 were not related, but then I remember reading somewhere the ESXi hosts need to have the same time in order for HA to work.

Regarding issue 1:

The 10 hosts have a difference of 5-6 minutes between them. Oddly enough, the first hosts that are joined to a HA cluster, if their time is similar, they wil work.

But the others are more than 1-2 minutes off, they will fail.Our NTP is configured to 2 internal NTP servers, that sync with the outside world - the 2

NTP servers are:

The default gateway of each ESXi

The Core Network Switch

We know the NTP servers work properly since our AD is syncing with them and we checked local time on our PDC with external time source.

Debugging information

Output of /etc/ntp.conf

~ # cat /etc/ntp.conf
restrict default kod nomodify notrap nopeer
restrict 127.0.0.1
server 10.*
server 10.*
driftfile /etc/ntp.drift

Everything was configured through the GUI, i'm showing the output of ntp.conf for confirmation.

We have tried rebooting hosts (no success).

Anyone else knows what we can do to troubleshoot this?

Thanks,

Ionut

[dkfbp]

Have you added your ESXi hosts to your windows domain? If that is the case the hosts stop syncing time with NTP but gets its time from the domain.

[DyJohnnY]

Hi,

No, not added to domain. Simple configuration.

[matuscak]

Does your ntp.conf literally contain "server 10.*"?

If so, that doesn't look right to me. The way I read the NTP documentation, that should be a DNS name or a IP 4/6 address.

If you have a system with the regular NTP distribution on it, you can do a:

ntpq -p NTPserver

and it should show you if it's synchronized or not.

[DyJohnnY]

Hi,

No it doesn't,I just did not want to include the FULL ip address of internal hosts on the web.

Those are IPv4 addresses of cisco devices.

In the meantime we seem to have solved our issue, which was related to a malfunctioning NTP service on the cisco switches(service was running but not responding as it should).

Now all our hosts are in sync.

[DSTAVERT · Accepted solution]

You have tried restarting NTP.