This is a discussion on Re: BIND 9.4.2P2 on Solaris 10 Sparc - do we have a problem? - DNS ; JINMEI Tatuya / ?$B?@L@C#:H writes: > At Fri, 22 Aug 2008 16:53:56 -0400 (Eastern Daylight Time), > Mike Diggins wrote: > > > There have been numerous postings from people having problems with BIND > > 9.5.0x and Solaris 10 ...
JINMEI Tatuya / ?$B?@L@C#:H writes:
> At Fri, 22 Aug 2008 16:53:56 -0400 (Eastern Daylight Time),
> Mike Diggins
> > There have been numerous postings from people having problems with BIND
> > 9.5.0x and Solaris 10 (too many open file descriptors). I see a few posts
> > of similar problems with 9.4.2x and Solaris 10. I plan to upgrade my
> > Solaris 10 SPARC, BIND 9.3.5P1 machines to 9.4.2-P2 next week. I'm a bit
> > confused as to whether this is a Solaris 10 INTEL problem, or it affects
> > SPARC as well. Am I going to run into this file descriptor problem?
> As far as I know it's architecture independent. If you see a 'too
> many file descriptor' errors with the default installation parameters,
> you may want to rebuild BIND9 with a larger FD_SETSIZE.
Running selecttest, I have built 9.4.2-P2 with FD_SETSIZE=4096 and deployed
it on our internal root servers running Solaris 9 SPARC. After a week of
monitoring fd usage (once a minute via cron) I found that it remained
way below the configured limit, and we haven't logged any "too many open
file descriptor" errors.
There were other errors I haven't seen before, though, like
dispatch: info: dispatch 1ecfb10: shutting down due to TCP receive error: 18.104.22.168#53: connection reset
general: error: malformed transaction: hosts/internal.com.jnl last serial 964384 != transaction first serial 964383