On Sep 29 2008, Mike Diggins wrote:

>Thanks for the reply. So instead of creating the zone from scratch and
>HUP'ing named, apply the static entries (from the database) using nsupdate
>(or similar)? I wouldn't mind having a look at your script if you don't


Bear in mind it has some parochial features; this isn't polished software.

>Assuming a database of 30,000 hosts with various DNS records (PTR, CNAME's
>etc) for each, how long would it take to nsupdate that?

Well, you wouldn't want to update all of them every time, because most
of them wouldn't have changed (one hopes). In any case, single DNS
updates (nsupdate's "send") are limited to 64KB over the wire.

As to how long "genupdates" takes to generate a small diff from two
zone files of that size, cam.ac.uk has about 43000 RRs and we can
live with it. Actually the preliminary "named-checkzone -D" to
normalise the new input takes longer than the "genupdates"[*].
You save the normalised form for comparison next time around,
of course.
[*] But in a previous life, "genupdates" tried to syntax-analyse zone
files in a more, although not entirely, general format, and it was *slow*.
${DEITY}, was I glad to be able to throw away that code!

Chris Thompson
Email: cet1@cam.ac.uk