bind 8.2.3 memory leak

[Mark.Andrews at nominum.com]

> Thanks for your reply Mark.
> 
> When I say all zones I mean all valid zones.  Invalid queries happen all the
> time (as I'm sure you're aware).  BTW, these name servers have no connection
> to the internet and are root servers.

	How can there be an "invalid" query as you claim these servers
	serve all of the name space.  All queries should fall within
	one of the zones being served (including ".").  If they don't
	then your claim is false and they are caching servers.

	Note BIND 8.2.3 will fail to serve a zone if it detects a error
	while loading and will behave as a caching server for the
	namespace that would otherwise be serving.

	Mark
> 
> I set max-ncache-ttl to 180 to limit the negative cache entries kept.
> 
> For reference, one nameserver has been running for 12 hours now and has gone
> from about 6 Mb up to about 17Mb.
> 
> Line from top ...
> 
>   PID USERNAME PRI NICE  SIZE   RES STATE   TIME   WCPU    CPU COMMAND
> 16494 root      33    0   17M   10M sleep  18:00  1.41%  1.40% named
> 
> It would be nice if named did stabilise, but we've had occurrences where
> named filled up 120Mb of swap space.
> 
> Thanks,
> 
> Peter Anderson
> Senior Communications Analyst
> Ph:  0011 61 2 99025938
> <Remove ETER from my address to reply>
> 
> Any views or opinions presented are solely those of the author and do
> not necessarily represent those of Westpac Banking Corporation.
> 
> 
> 
> 
> -----Original Message-----
> From: Mark.Andrews at nominum.com [mailto:Mark.Andrews at nominum.com]
> Sent: Wednesday, April 04, 2001 1:31 PM
> To: ANDERSON, Peter
> Cc: 'comp-protocols-dns-bind at moderators.isc.org'
> Subject: Re: bind 8.2.3 memory leak 
> 
> 
> 
> > Hi,
> > 
> > I believe 8.2.3 has a memory leak.
> > 
> > I've got no idea what routine is causing it, but has anyone else seen the
> > same thing?
> > 
> > I've tried to limit the amount of memory being used with the datasize
> > option.
> 
> 	This is not its purpose.  Its purpose is to allow the server
> 	to request more than the default per-process memory allowance
> 	from the system.
> 
> > 
> > This DOES seem to work (memory used doesn't get over the limit) but named
> > still core dumps.  Is this what it's supposed to do?
> 
> 	Yes.
> 
> > 
> > We have 6 internal DNS servers all authoritive for all zones.  The only
> > increase in data size should be from negative cache entries.
> 
> 	If they are authoratative for *all* zones then they won't need to
> 	cache anything.
> 
> > 
> > I've set max-ncache-ttl to 180.
> > 
> > After initial startup named uses between 5 & 6 Mb RAM.  Over time this
> grows
> > (faster with more usage) until the swap space fills (or datasize is hit).
> 
> 	This does not sound like an authoratative server.  It sounds like
> 	a caching server.
> 
> > 
> > Can anyone help?
> 
> 	See max_cache_ttl (src), cleaning-interval and max-ncache-ttl.
> 	As shipped named is designed to stabalise its memory usage
> 	after 7 days assuming a consistant query distribution.
> 
> 	Mark
> 
> > 
> > Thanks,
> > 
> > Peter Anderson
> > Senior Communications Analyst
> > Ph:  0011 61 2 99025938
> > <Remove ETER from my address to reply>
> > 
> > Any views or opinions presented are solely those of the author and do not
> > necessarily represent those of Westpac Banking Corporation.
> > 
> --
> Mark Andrews, Nominum Inc.
> 1 Seymour St., Dundas Valley, NSW 2117, Australia
> PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at nominum.com
--
Mark Andrews, Nominum Inc.
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at nominum.com