RE: Settings for File Descriptors

[Hayward, Bruce]

Thanks much

Email to bruce.hayw...@mtsallstream.com


> -Original Message-
> From: Chris Thompson [mailto:c...@hermes.cam.ac.uk] On Behalf 
> Of Chris Thompson
> Sent: Friday, March 06, 2009 12:12 PM
> To: Hayward, Bruce
> Cc: Bind Users Mailing List
> Subject: Re: Settings for File Descriptors
> 
> On Mar 6 2009, Hayward, Bruce wrote:
> 
> >I am trying to understand file descriptors with bind in 
> mind, and what 
> >they should be set at in conjunction with the OS.
> > 
> >We are running 9.5.1 P1 on Solaris 10 (patched up), which is 
> basically 
> >all that is on each server.
> > 
> >Some questions:
> > 
> >1) Is there a recommended setting (number of FDs)?
> 
> BIND will set the file descriptor resource limit to 
> "unlimited" when it starts up, unless you tell it not to in 
> the "options" setting.
> Just forget about it.
> 
> >2) Better to set this at the kernal level with the set rlimit, or a 
> >compile option with configure, or both
> 
> This doesn't apply to BIND (see above) but I think it foolish 
> to mess around with /etc/system to change resource limits for 
> all processes, when you can easily use a targetted "ulimit" 
> command in the calling script for the process that actually needs it.
> 
> BIND doesn't have a compile option for this, anyway, AFAIK. 
> Are you confusing this with the FD_SETSIZE value? BIND 
> 9.5.1-P1 on Solaris will not be using select(3c) anyway, so 
> that isn't an issue.
> 
> >3) Should the number be correlated to the number of 
> recursive clients?
> >4) Different on the auth servers?
> 
> If you did go out of your way to make it other than 
> "unlimited", then yes to both. But don't.
> 
> --
> Chris Thompson
> Email: c...@cam.ac.uk
> 
___
bind-users mailing list
bind-users@lists.isc.org
https://lists.isc.org/mailman/listinfo/bind-users

Re: Settings for File Descriptors

[Chris Thompson]

On Mar 6 2009, Hayward, Bruce wrote:


I am trying to understand file descriptors with bind in mind, and what
they should be set at in conjunction with the OS.

We are running 9.5.1 P1 on Solaris 10 (patched up), which is basically
all that is on each server.

Some questions:

1) Is there a recommended setting (number of FDs)?


BIND will set the file descriptor resource limit to "unlimited" when
it starts up, unless you tell it not to in the "options" setting.
Just forget about it.


2) Better to set this at the kernal level with the set rlimit, or a
compile option with configure, or both


This doesn't apply to BIND (see above) but I think it foolish to mess
around with /etc/system to change resource limits for all processes,
when you can easily use a targetted "ulimit" command in the calling 
script for the process that actually needs it.


BIND doesn't have a compile option for this, anyway, AFAIK. Are you
confusing this with the FD_SETSIZE value? BIND 9.5.1-P1 on Solaris
will not be using select(3c) anyway, so that isn't an issue.


3) Should the number be correlated to the number of recursive clients?
4) Different on the auth servers?


If you did go out of your way to make it other than "unlimited",
then yes to both. But don't.

--
Chris Thompson
Email: c...@cam.ac.uk
___
bind-users mailing list
bind-users@lists.isc.org
https://lists.isc.org/mailman/listinfo/bind-users