This same NTP server is a virtual machine hosted in Azure and remained the entire year with fixed score at 20, but since day 22 the score began to be reduced due to the high offset.
Due to the behavior of the monitoring station graph, the score appears to be gradually reduced until the NTP service resynchronizes the time with the time source, as if something were causing the clock to be delayed in a short period of time.
I’m using three distinct stratum 1 time sources.
The test result made from another server using the ntpdate -q command shows an offset of -0.034593.
I also use a DNS and HTTP services monitoring application called Uptimerobot, which has shown a considerable increase in latency of requests in recent days.
Could anyone tell me what I can do to try to solve this? I haven’t made any modifications to the server recently.
The three stratum 1’s you’re using are working fine.
My NTP monitoring client shows the same behavior as the NTP Pool monitor.
One guess is that the VM is resource starved. Try increasing the polling rate. e.g.,
server a.b.c.d minpoll 7 maxpoll 7
I think other virtual machines present on the same host have gone on to consume a lot of resources in recent days, increasing the clock delay in short time.
Alternatively, I stopped using time synchronization with external NTP sources and configured my virtual machine to sync directly with host time. According to Microsoft documentation, “Azure hosts are synchronized to internal Microsoft time servers that take their time from Microsoft-owned Stratum 1 devices, with GPS antennas. Virtual machines in Azure can either depend on their host to pass the accurate time ( host time ) on to the VM.”
After I started using host time sync instead of using external sources, I noticed that the clock offset has decreased and is much more stable. The only downside is that now my server is considered a Stratum 4.
Is there a problem using a server that synchronizes with host time in NTP Pool instead of external time sources?
The time is stable now and seems accurate. If one trusts the Microsoft documentation and implementation, that should be sufficient. While I have little direct experience, I do know that Microsoft invested a lot of effort into this.
Not at all. The host itself synchronizes with other servers. Besides, this is not different from many public servers that some colleges and companies make available for general use, whether in the pool or not.