Symptom
You add or delete machines or servers but your changes are not recognized or do not take effect. Or in some instances the changes are recognized and at other times they are not yet in effect.
Probable cause
The most likely cause is that you forgot to increment the SOA serial number on the master server after you made your change. Since there is no new SOA number, your slave servers do not update their data to match that of the master so they are working with the old, unchanged data files.
Another possible cause is that the SOA serial number in one or more of the master data files was set to a value lower than the corresponding serial number on your slave servers. This could happen, for example, if you deleted a file on the master and then recreated it from scratch using an input file of some sort.
A third possible cause is that you forgot to send a HUP signal to the master server after making changes to the primary's data files.
Diagnosis and solution
First, check the SOA serial numbers in the data file that you changed and the corresponding file on the slave server.
If the SOA serial number in the master file is equal to, or less than, the serial number in the slave file, increase the serial number on the primary's file so that it is greater than the number in the slave file. For example, if the SOA number in both files is 37, change the number in the primary's file to 38. The next time the slave checks with the primary, it will load the new data. (There are utilities that can force a master to immediately transfer data to the secondaries, if you have one of these utilities you can update the slave without waiting for it to check the primary.)
Review the syslog output for the most recent named nnnn restarted or named nnn reloading nameserver entry. If the timestamp for that entry is before the time you finished making changes to the file, either reboot the server or force it to read the new data as explained in How to Force in.named to Reload DNS Data.