Time server when deploying disconnected ASDK

Question

When deploying a disconnected ASDK, it is recommended install the ASDK while connected to the internet.
However for the time server IP, is it ok to use internet time server IP during deployment?

Would it be necessary for this time server to be connectable after disconnecting the ASDK from internet?
If yes, what is the workaround to connect to local time server if only one network adaptor which is connected to internet are used during deployment.

Answer 1

Hi @darkbluelion ,

once your ASDK is disconnected from the Internet, AzS-DC01 VM will revert to its (virtual) CMOS clock as timesource. Get-AzSTimeSource run into a PEP session will prove that. Namely, "type", normally set to NTP for DC01 will show CMOS.

ComputerName : ****-DC01
DomainRole : Domain Controller
Service : Running
Type : NTP <<<< this will change to CMOS
NtpServer : . . .
. . .

I am not sure if the virtual CMOS clock is fed by the underlying host at some point. For sure it will if DC01 is rebooted. And the physical host, being part of the domain itself, will sync with DC01, so there might be a certain degree of circularity.

However, in practice, you won't notice particular disruptions but it's likely your entire Azure Stack Hub will observe a drift of its own time over time.

To prevent it from drifting you need to set a local NTP service as recommended. It can be done using Set-AzSTimeSource in your PEP session, after installation if you want to replace the initial one set during installation.
Please be aware in many cases you need to set it using ",0x8" (no quotes) if the NTP server isn't Windows.

More on Set-AzsTimeSource and the ,0x8 switch here: https://learn.microsoft.com/en-us/azure-stack/operator/azure-stack-configure-time?view=azs-2005

Best regards,
Fabio