Bind 9.3.3 on Solaris 9

Chris Thompson cet1 at hermes.cam.ac.uk
Fri Dec 22 19:22:46 UTC 2006


I have been hoping that someone else would respond to this posting of
10 days ago, as I would be quite interetsed in tuning hints for a
similar environment ...

On Dec 12 2006, vince anton wrote:

>Hi All,
>Im in the process of setting up bind 9.3.3 on a sparc solaris9 box with 1G
>RAM and 2 CPUs, but may add more RAM and CPU if needed. This will be for
>recursive resolution for my users, as well as as a slave for domains that I
>host.

In our case: we've just upgraded our central namesevers to 2 Sun V60s each
with 4 ("virtual") processors and 2GB, and they are running Solaris 10 (x86).
Each nameserver instance runs in its own Solaris 10 "zone" - eventually we 
will be running one recursive instance and one authoritative instance on 
each box (and from time to time a testing one as well...). BIND version is
currently 9.3.3.

>Ive been reading through various posts to ensure that I avoid issues with
>slow resolutions etc...  and ensure that the box will handle moderate
>request load.
>
>So after looking through the archives and googling around, Ive gathered some
>info, but still need some stuff claririfed by anyone running on Solaris 9:
>
>1. There seem to have been issues when compiling multithreaded versions of
>bind, however this seems to be mainly on FreeBSD and linux, and not on
>Solaris ?
>
>2. Also some people seem to have had better performance with named when
>setting ISC_MEM_USE_INTERNAL_MALLOC  to 1.  Is this also true for a solaris
>box with 2 or more CPUs ?

That's a question I would very much like to see answered by someone with
the relevant experience.

>So for solaris 9, what are people out there doing - is there a need to apply
>some tweaks such as the ones above, or disable threading or whatever at
>compile time, or tune to the OS kernel variables to optimise  performance ?

We've been running multi-threaded BINDs on Solaris 8 and Solaris 9 for some
time (without obvious problems) but mostly on uni-processors until now.

-- 
Chris Thompson
Email: cet1 at cam.ac.uk



More information about the bind-users mailing list