question on retry and notifications in v8.2.3

[Kevin Darcy]

Thomas Duterme wrote:

> Just looking for a confirmation on the following.
>
> If I update a master server and send a SIGHUP to named, it will automatically send out notify messages to the defined slaves after a few seconds.
>
> If so, then what is the point of the refresh value in the SOA?

Maybe you have "stealth" slaves.

Maybe you even have "super-stealth" slaves, i.e. slaves that nobody told you about and whose existence you can only ascertain from the query logs. If you don't know about a slave, you can't add it to the also-notify list.

Or maybe you want to keep the also-notify list down to reasonable size to avoid thrashing the master server on restarts and reloads. Sometimes it's not necessary for *every* slave to have a fast convergence, only the "central" slaves.

> A secondary name server would never have to use this since the master would send out notification on any changes, right?  Or is this something left in to make sure that the slave would update regardless of whether or not the master broadcast message was heard?  (ie in case of network problems, does the master continue to send out multiple notifies, or does it act like the elitist that it is and essentially say 'to hell with the slaves, its beneath me to notify them more than once.')

BIND's implementation of NOTIFY will attempt retries, but eventually gives up.

Um, by the way, it's not really a "broadcast" (or even multicast) message in the technical sense...

                                                                                                                - Kevin