My ancient ntp server had to go a while back due to its size and power consumption. In recent days I bought a new modern ntp server which is GPS+PPS bound but no heater or stabilised crystal.
My new ntp server (an ebay bargain) does not offer its own fallback external servers, so I am configuring them on my router. However, I have run into a few snags in the effort to configure a genuine fallback-only external ntp set. For routine use I do not wish for any external ntp traffic unless the internal one has failed.
In the past I could achieve my aim by using the "true" statement on my own ntp, which pinned the snowflake on the correct address and the external providers remained dormant. Back then the "prefer" statement did little to prevent the snowflake moving to the best "+" server and blending the results.
The "True" statement doesn't seem to work (at least for my Debian router) - it accepts the statement but nothing seems to change and all servers, internal & external, remain in regular use. I'm not sure if the absence of a RTC in my router is influencing the "true" statement or not.
In better news the "prefer" statement now works perfectly and pins the snowflake as intended and the stratum 1 pps is the master clock. Great for accuracy and redundancy but still has ntp packets traversing the WAN when not needed or desired.
My own ntp server snowflaked (partially redacted address) along with the other curated/resilient external stratum 1 mix, feeding my router and my wider network:
The rather sparse ntp status once trained and settled:
NTP Server Configuration (10.0.1.50 is my internal ntp):
Any thoughts or settings I could try to achieve fallback-on-fail-only on my ntp sources and keep my internal ntp as the sole server at all other times?
My new ntp server (an ebay bargain) does not offer its own fallback external servers, so I am configuring them on my router. However, I have run into a few snags in the effort to configure a genuine fallback-only external ntp set. For routine use I do not wish for any external ntp traffic unless the internal one has failed.
In the past I could achieve my aim by using the "true" statement on my own ntp, which pinned the snowflake on the correct address and the external providers remained dormant. Back then the "prefer" statement did little to prevent the snowflake moving to the best "+" server and blending the results.
The "True" statement doesn't seem to work (at least for my Debian router) - it accepts the statement but nothing seems to change and all servers, internal & external, remain in regular use. I'm not sure if the absence of a RTC in my router is influencing the "true" statement or not.
In better news the "prefer" statement now works perfectly and pins the snowflake as intended and the stratum 1 pps is the master clock. Great for accuracy and redundancy but still has ntp packets traversing the WAN when not needed or desired.
My own ntp server snowflaked (partially redacted address) along with the other curated/resilient external stratum 1 mix, feeding my router and my wider network:
The rather sparse ntp status once trained and settled:
NTP Server Configuration (10.0.1.50 is my internal ntp):
Code:
name-server 127.0.0.1
ntp {
server 10.0.1.50 {
prefer
}
server ntp.se {
}
server time.euro.apple.com {
}
server time.google.com {
}
Any thoughts or settings I could try to achieve fallback-on-fail-only on my ntp sources and keep my internal ntp as the sole server at all other times?