Best Practice for Time Sync for Active Directory Domain Controllers

scottalanmiller

OK, I have VM Tools running - do I just wait and see?

There is no sync involved. You are mixing the concepts of NTP/SNTP with a source server and actually controlling the system's clock.

scottalanmiller

@Dashrender said:

@scottalanmiller said:
@Dashrender said:
OK tried a
 w32tm /resync
and got back
 The computer did not resync because no time data was available.
What time source do you have set? w32tm requires an SNTP source to sync to, what SNTP server do you have it talking to?
I don't, it's currently pulling from
 Local CMOS Clock

Right, which is unrelated to time syncing.

Dashrender

I'm lost
I said I did this.

@Dashrender said:

My current PDC emulator is set to pull time from the BIOS clock
 C:\Windows\system32>w32tm /query /source
 Local CMOS Clock
This hasn't been an issue for years, yet someone called this morning and reported that the phones and the computers didn't match timewise, so I'm looking into it.

So my PDC emulator is pulling time from CMOS. If CMOS changes, won't the OS change too?

Dashrender

Shouldn't this be in IT discussions? It's technical in nature.

Dashrender

I guess my desire here was to have the ESXi host be the main source for time inside my network. It of course would pull time from the internet.

It sounds like this isn't going to work. So instead I have to have my PDC emulator pull it's own time from the internet, and the VM Hosts will have to be managed separately.

Dashrender

Here's the command to set your PDC emulator to sync with a time source

  w32tm /config /manualpeerlist: peers /syncfromflags:manual /reliable:yes /update

Replace peers with your FQDN or IP of the desired time servers.

scottalanmiller

@Dashrender said:

Shouldn't this be in IT discussions? It's technical in nature.

Hmmm... I didn't choose the category, it just did it.

scottalanmiller

@Dashrender said:

I guess my desire here was to have the ESXi host be the main source for time inside my network. It of course would pull time from the internet.

It sounds like this isn't going to work. So instead I have to have my PDC emulator pull it's own time from the internet, and the VM Hosts will have to be managed separately.

ESXi can pull time from the Internet. If it is correct and the DC is getting its time from the ESXi clock then the ESXi is setting the DC which, in turn, uses SNTP to talk to the rest of the network.

scottalanmiller

Here is vmware's older paper on how they recommend that this be set up:

http://www.vmware.com/files/pdf/Virtualizing_Windows_Active_Directory.pdf

scottalanmiller

Here is a more recent one:

http://www.vmware.com/files/pdf/solutions/Virtualizing-Active-Directory-Domain-Services-on-VMware-vSphere.pdf

scottalanmiller

VMware definitely recommends that you use an external time source to control drift, not using the ESXi virtualized clock.

dafyre

I would set ESXi host to use $external_NTP... and then point the DCs to $external_NTP and then all of the clients will magically sync with DCs.

Jason

You set your PDC Emulator to pull from a reliable NTP server then the rest will sync from that ex:

w32tm.exe /config /manualpeerlist:”0.us.pool.ntp.org 1.us.pool.ntp.org 2.us.pool.ntp.org 3.us.pool.ntp.org” /syncfromflags:manual /reliable:YES /update

scottalanmiller

I just heard to the tune of "Free your mind, and the rest will follow..." in my head:

Sync your time, and the rest will follow...

Dashrender

Alright - I read through the document that Scott provided about VMWare and time syncing.

The reason VMWare wasn't changing my PDC emulator's clock was that time syncing between ESXi and the VM was disabled (default behavior).

Edit the VM session, Click on the Options tab, click on VMware Tools and you'll see these two check boxes at the bottom on the right.

Make your desired choices, save and you're good.

Dashrender

@scottalanmiller said:

VMware definitely recommends that you use an external time source to control drift, not using the ESXi virtualized clock.

I wouldn't ever rely solely on their virtual clock, I'd definitely like ESXi itself to be syncing to something.

Dashrender

@dafyre said:

I would set ESXi host to use $external_NTP... and then point the DCs to $external_NTP and then all of the clients will magically sync with DCs.

With concerns about Windows and Time, the only server that you should have syncing with an outside source is the PDC emulator. All other domain devices will sync from that machine.

scottalanmiller

@Dashrender said:

@scottalanmiller said:

VMware definitely recommends that you use an external time source to control drift, not using the ESXi virtualized clock.

I wouldn't ever rely solely on their virtual clock, I'd definitely like ESXi itself to be syncing to something.

Of course, no clock anywhere just relies on itself!

scottalanmiller

@Dashrender said:

@dafyre said:

I would set ESXi host to use $external_NTP... and then point the DCs to $external_NTP and then all of the clients will magically sync with DCs.

With concerns about Windows and Time, the only server that you should have syncing with an outside source is the PDC emulator. All other domain devices will sync from that machine.

Only if the PDC emulator is using NTP. If it is using the local clock then the hypervisor has to fulfill that role.

Dashrender

@scottalanmiller said:

@Dashrender said:

@dafyre said:

I would set ESXi host to use $external_NTP... and then point the DCs to $external_NTP and then all of the clients will magically sync with DCs.

With concerns about Windows and Time, the only server that you should have syncing with an outside source is the PDC emulator. All other domain devices will sync from that machine.

Only if the PDC emulator is using NTP. If it is using the local clock then the hypervisor has to fulfill that role.

I did say Windows and Time. If the PDC emulator is using the local clock, that local clock would be Windows outside source, but if that's all you're doing, then definitely you should be syncing the local clock (ESXI, Hyper-V XenServer, etc) with an atomic source if possible.