I have a 2 vm instances in GCE that do not keep correct time in spite of the server OS being set to sync with NTP every 24 hours. The OS clock gains about 1 second of time every four minutes (or about 15 seconds each hour). One vm instance machine type is n1-standard-8 (8 vCPUs, 30 GB memory) and the other is Custom(24 vCPUs, 128 GB memory) both are in us-cental1-a timezone and are shown by Compute Engine as being "under utilized".
The Time for eacg is set to sync each 24 hours to time.nist.gov.
We have noticed that when the time get's off more than 5 minutes it has caused some issues with our other processes.
These vm's are 2 among several VM instances running in Google Compute Engine and are fully patched (windows 2008 r2 standard SP1). These 2 vms' are the only ones that seem to be having this time issue.
If anyone could provide some direction on this i'd be so very appreciative. Thanks!
extra info: These instances were cloned into Google Cloud using the cloudendure migration service about a year and a half ago (along with several others that do not have this issue).
Upgrade?
First - you notice that 5 minutes off cause problems. Any admin will tell you that this is because Kerberos hates ticket time offsets more than that on windows ;)
Your main problem is server2008 - 2012 upward are desigend to be time stable, also if the host supports injecting the time into the client (clients generally should not have to rely on the network to get their time). Even if, then, 24 hours is simply too long a time - sync more often. Every 5 minutes, for example.
I had the same at the old times of using 2008 - now on 2019 we are super stable, but we also run 2019 on the host.