5

I have a old bind dns server (linux centos5.5 with bind 9.3.6) with 2000+ zone files. Whenever any change is done in any file the dns is restarted. I guess a reload must be sufficient though.

Problem is that a restart takes a long time 50s-60s and dns requests fail during that time.

This must be a very common issue , What is the correct way of restarting bind

8

You could reload just the specific zone that was changed:

rndc reload zonename

rather than restarting the whole server. That's the simplest way. I actually do something different on my production DNS:

  • Keep all my masters on one separate server (a tiny VM) that services NO user queries
  • Use 2 slave servers to service all queries that get all their zones replicated from the master
  • All changes get made on master, if a reload is needed only the affected zones get retransferred to the slaves.
  • If a reconfig is required (ie changing named.conf to add/delete zones or whatever) schedule that for a maintenance window.
  • 1
    I've always preferred to have the master copy of your zones on a different server than the one serving your clients. Additionally, and especially on a serving master, you should be checking any changes with named-checkconf and named-checkzone before you reload the zone. – Rick Buford Sep 5 '15 at 16:28
  • Skip the filename, i.e. rndc reload, to reload the whole thing. – Ethan Furman Aug 17 '18 at 22:35
2

service named reload should also reload and not restart bind. I agree with others for maximum availability have separate slave bind servers which can serve dns requests if you do need to do maintenance on your master dns server.

  • the service seems to be 'named-chroot' on my server, not sure if that is a customisation – nick fox Aug 24 '17 at 11:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.