I'm currently in the process of migrating my BIND master host to a new physical machine and IP. I've installed on the new master and tested out updating a zone file and seeing the slaves take the new serial. I've also updated /etc/resolv.conf in my production environment to ensure no hosts are hitting the old master...however, checking the query logs I still see lots of queries hitting the old BIND master (mostly during puppet runs). Trying to play whack-a-mole with restarting services dependent on querying will take too long...I have a few questions:
- Anyone familiar with this problem with nameserver caching? If so, how did you resolve it?
- Would shutting off named on the old master resolve this problem?
- In the meantime, would having admins update the serial and reload named in two places be bad for the slaves?