Bind 8.1.2 in.named memory leak in Solaris 7

[jgomide at bancobrasil.com.br]

Hi Jim

When I start named, it takes 4M. After 3 hours it is eating 16 M of memory.
About the bind version, bind 8.1.2 in Solaris 7 (message subject). I believe
this version *has* a memory leak.
I´m sure I have to upgrade to bind 8.2, but where do I get it compiled for
Solaris 7?

[]s

Joaquim Gomide

--
   /\//  Joaquim Eudes Mendes Gomide
  / /\   Analista de Informatica
  \/ /   Banco do Brasil S.A.   http://www.bancobrasil.com.br
  / /\   Fone: (061) 310-6303 (Brasil: +55)  Fax: (061) 310-6386
  \/ /   mailto:gomide at bancobrasil.com.br
 //\/    I am on my own. It´s not Banco do Brasil´s opinion.





Jim Reid <jim at mpn.cp.philips.com> on 08/10/99 12:04:38

To:   F5028449 Joaquim E M Gomide/BANCO DO BRASIL at bancobrasil.com.br
cc:   bind-users at isc.org

Subject:  Re: Bind 8.1.2 in.named memory leak in Solaris 7




    Joaquim> I´m having the  problem describe in the article. Look at
    Joaquim> top´s information below:
    Joaquim>  PID USERNAME THR PRI NICE  SIZE   RES STATE   TIME    CPU COMMAND
    Joaquim> 28729 root       1  58    0  545M  417M sleep  33:57  2.78%
in.named
    Joaquim> 29065 root       1  58    0 2120K 1504K cpu     0:00  0.51% top

Wow! A name server of that size is unusual. You might expect something
to be that big at a major ISP, but not for the environment you
describe.

    Joaquim> The server is just 3 day up and it is only a name server.
    Joaquim> This is a internal dns server and is used for zone
    Joaquim> transfer, so the cache should not grow as big
    Joaquim> as it is. We don´t have so many internal hosts to cache.

Hmm. There are a few things you could check. The first is the name
server's config file and zone files. Are you *sure* these are really
as small as you say they are? How big is named after it has been
started and it has only loaded your zone files? The second thing to do
would be to make the name server dump its cache - assuming you have
enough free disk space for it somewhere. That would let you find out
what was in the cache: maybe there's some idiot application that's
flooding the name server with queries, perhaps doing an exhaustive
search of the name space. ie
     for (i = a; i <= zzzzzzzzzzzzzzzzzzzzzzzzzzz; i++)
          lookup (i.com);

A third possibility could be you're running really ancient name server
code that has a memory leak. [You didn't say which version of BIND you
were running.] Try running 8.2.1. It's always a good idea to run up to
date name server code because it should have fewer problems with
memory leaks (memory that doesn't get freed whenever it is no longer
needed). It'll also have the known security holes plugged.