Currently i am deploying a network wide NTP set up. All works fine but we have some confusion over the use of " NTP master "
If we do not include the NTP master command on the catalyst cores the NTP set-up does exactly what we expect.
When we add NTP master on the core catalyst switches with a high startum ( ntp master 5 for example ) we see that the cores seems to flick between the internal clock on 127.x.x.x and the actual main NTP source server on the network - even though the stratum value of this server is lower ( ie stratum 3 )
than the actual core NTP master 5 .....
Is this normal behaviour ? does the catalst core look at all the clocks and decide its the best time source regardless of stratum value ?? does it occasionally look at clocks and then decide that the server is sometimes the best source ( when we do sh ntp associations the NTP master seems to change from the core switch to the true NTP master server and then back again ) ?
Or should we simply not bother with the NTP master command - I understand ( maybe mistakenly ) that the master command just means that the cat will use itself as master when it loses connection to the main NTP server ?
any help gladly appreciated.
It could be a timing issue.
The NTP server with stratum of 3 does not respond on time the switch automatically looks to itself as a time provider.
You should remove the 'NTP Master' command to avoid this issue. By default, if this switch is the only one configured with the external NTP server, it will act as NTP Master for the rest of devices in your network pointing to this switch as the NTP server.