Logging of rate-limited queries way too talkative

[Stephane Bortzmeyer]

I'm trying RRL on the new BIND 9.9.4.

When RRL steps in, if I understand the documentation properly, two
things are logged, a summary of the beginning and end of RRL, and one
message per rejected query (!) Since RRL is used when there is an
attack, there are *many* such messages. Worse, the default behavior of
BIND aggravates the attack by filling the disk.

>From the ARM, I find no way to prevent these annoying messages,
except shutting down all query errors logging. Here is my best config
so far:

logging { 
   channel rrl_channel {
      file "rrl.log" versions 3 size 10m;
      print-time yes;
      print-category yes;
   };

   # Prints begin and end of rate-limiting
   category rate-limit {
      rrl_channel;
   };

   # Individual requests rejected
   category query-errors {
      null;
   };

};

But it is too harsh, since all query errors are sent to null.