Upgrade challenges

[Castricum, Ben (W&O SO OPS NET Netwerk beheer)]

> -----Original Message-----
> From: Jinmei_Tatuya at isc.org [mailto:Jinmei_Tatuya at isc.org] 
> Sent: dinsdag 15 juli 2008 19:47
> To: Castricum, Ben (W&O SO OPS NET Netwerk beheer)
> Cc: bind-users at isc.org
> Subject: Re: Upgrade challenges
> 
> At Tue, 15 Jul 2008 12:13:25 +0200,
> "Castricum, Ben (W&O SO OPS NET Netwerk beheer)" 
> <ben.castricum at kpn.com> wrote:
> 
> > I just gave 9.4.3b2 a try while setting FD_SETSIZE to 32768. First
> > attempt resulted in
> 
> If you use /dev/poll, the setting of FD_SETSIZE doesn't matter.
> 
> > named[599]: main.c:502: unexpected error:
> > named[599]: isc_socketmgr_create() failed: file not found
> > named[599]: create_managers() failed: unexpected error
> > named[599]: exiting (due to early fatal error)
> > 
> > but this was caused by /dev/poll not available in the chrooted
> > environment (at least we know it uses poll as API). After 
> creating this
> > device, bind seems to start correctly. However after a few 
> seconds the
> > logfile gets flooded with "*** POKED TIMER ***" messages, several
> > thousands per second.
> 
> Which operating system and version are you using.  I guess it's
> Solaris, but I'd like to be sure about it, and would like to know the
> OS version.  How many CPUs are available on your machine, and how many
> threads is named using?

I should have mentioned this in my previous post, but I am using Solaris
8 with 2 US-IIIi CPUs.

> If you disable threads, does that provide
> reasonable performance without other problem?

I recompiled with --disable-threads and all seemed well (no noticeable
performance loss). However when I switched again to the threaded version
I did not get the POKED TIMER message again. I have been closely
monitoring the logfiles for several hours and all seems fine. Named uses
7 threads now. 

Thanks for your support,
Ben