don't hate the messenger

[Bill Manning]

% 
% >>>>> "Thomas" == Thomas Duterme <thomas at madeforchina.com> writes:
% 
%     Thomas> Why are SIGHUPs *bad* as a way to restarting named?
% 
% Because it's crude. If all that's changed is one zone, why reload (not
% restart BTW) the whole server? And maybe you shouldn't have to be the
% UID running the name server to manage it? With ndc and rndc, that's
% possible. It's not if you send signals. Also, the signals used for
% "controlling" the name server have changed from time to time. And in
% BIND9, most signals that get sent to the server will make it
% terminate. So getting into the habit of using a proper control tool
% that doesn't use signals to talk to the name server is a good idea.
% 

Ah, but the tools for fine grained control don't always work the way you want 
them too. Denegrating a tool in the toolbox 'cause its a Really Big Hammer
is ... (nice words about social engineering placed here)

For example:

[@z named] rndc stop
rndc: connect: connection refused
[@z named] rndc status
rndc: connect: connection refused
[@z named] ps -ax | grep named
10068 ?        S      0:00 named
10069 ?        S      0:00 named
10070 ?        S     63:21 named
10071 ?        S      0:00 named
10072 ?        S      1:26 named
18228 ttyp0    S      0:00 grep named
[@z named] kill -9 10068

Sometimes that hammer is needed. And for "small" servers with small zones, 
blowing the process away may not be such a big deal. And restarting the
process cleans up most of those unfree'd memory allocations... :)


--bill