Windows Time Service (NTP) should synchronize more frequently
Clocks on role instances drift too much in the 1 week interval between synchronizing clocks (http://blog.codingoutloud.com/tag/network-time-protocol/). I have observed differences of 2s on my own hosts, and this makes it difficult to use a time stamp in a database to (for example) indicate when a job on a worker role is expected to complete.
Up to 2s should be handled as acceptable drift. You can customize the time synchronization to use a different NTP server or change the internal in the startup tasks for your web/worker roles.
2 comments
-
Ruben Gonzalez
commented
Is there any update on this now that Windows Server 2016 has a more precise Windows Time Service?
-
Anonymous
commented
I don't see how such a performance is compatible of reliable time stamp for connection attempts linked to security in the frame of PCI DSS level 1 qualification claimed by Azure. Could you explicit what performances are now in 2015-01?