In article ,
"Walkenhorst, Benjamin" wrote:

> Hello everyone,
>
> In what respect is BIND's behaviour affected by the value of
> 'cleaning-interval'?
> BIND will look through its cache every minutes (default:
> 60) for expired entries and, if it finds any, removes them, that much I could
> find out in the manual.
>
> But what happens when an already expired entry is requested again, before
> being removed from cache? Will the server remove it just then (and probably
> re-query the RR)? This seems like the only sane behaviour I could come up
> with for this situation, so ... could someone confirm to me BIND does behave
> like this? (The alternative looks like passing expired RRs to resolvers which
> just seems silly).


Yes, that's what it does. In fact, in early versions of BIND this was
the *only* way that expired entries would be cleaned up.

> In what respect will modifying this value change the behaviour of the system?
> Will it just make BIND free up (de facto unused) memory faster (or slower,
> depending on how I change the value) or are there other effects to be
> expected?


That's all it does.

> If I am not entirely mistaken, the machines BIND runs on have... sufficient
> memory at least, so just _saving_ some won't be a big deal. On the other
> hand, we have to use BIND 9.1.1 right now which apparently doesn't implement
> a 'max-cache-size' option. So with that unavailable, could I use the
> 'cleaning-interval' for controlling - or at least influencing - the memory
> usage of BIND?
> Would setting this value too low (say, 1) adversely influence BIND's
> performance? (Like if it has a big cache, it might have to go all over it
> again and again...)


Yes. You'll be wasting lots of time, since the cache won't have grown
very much, and few entries will have expired since the last cleanup.

--
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***