Detecting Failures

[Martin McCormick]

	In the old days, a catastrophic failure of the master dns was
not nearly as devistating as it might be today in an Active Directory
environment.  Microsoft has a multimastering scheme where by several
platforms all behave as the master dns so that if one fails, nobody
notices outside of the system administrators, maybe.

	It is pretty easy to turn a slave in to a master as long as it
had all the zones to begin with.  That along with a second Ethernet
interface means that one just brings it up on the master's address
with the correct configuration file and things are able to be updated
again.

	This brings me to the question.  What do large operations with
extremely high reliability factors do when they run bind?

	Active Directory is growing fast in our group and we should
have a plan of action to quickly failover a dead master.

	The hard part to me seems to be how best to automate the
detection of a true dns failure in such a way as not to accidentally
trigger the switchover and possibly have two working systems sharing
the same IP address.

Martin McCormick WB5AGZ  Stillwater, OK 
OSU Information Technology Division Network Operations Group