Re: [Help-gnu-radius] Long startup time

[Greg G]

Gerald wrote:

> On Mon, 24 Nov 2003, Greg G wrote:
>
>  
>
> > This is on a Solaris 2.7 box.
> >    
>
> I thought everything was slow on a Solaris box? ;-)
>  
  Sure, but this is a wee bit excessive. :-()

> > When I start up (or HUP the radiusd), there's about a 2 to 3 minute
> > period between when I get the "users reloaded" message and the "Ready to
> > process requests" message.  truss shows me *nothing* during this time.
> > ps tells me that its' in the 'run' state or taking up 40%+ of one of my
> > CPUs.  What is it doing during this time, and is there possibly a way to
> > shorten it?  I don't want my radius server to be down that long if I'm
> > just HUPping it.
> >    
>
> Technically you are supposed to do radctl reload to reread config files. I
> think it currently just HUPs the process, but in case that changes in the
> future you might want to learn the correct way to start with.
>  
  Yeah, it currently just hups the thing.  I didn't even realize there 
was a ctl script.  I never look in /usr/local/sbin...

> Are any of the partitions that radius is interacting with using NFS?
> Either the logging partition or where you have your config files?
  Nope.

> Is DNS configured for forward and reverse of your primary IP address on
> this machine?
>  

  Yes.  Does that matter?  What is it doing during these two minutes?

> What version of gnu-radius are you using?
  1.1.  That's what I found as the latest version.

> > So, is this normal?
> >    
>
> It's not normal on my machines, but I'm also not on solaris.
>
> radius#: time radctl reload
> Reloading configs
>
> real    0m0.025s
> user    0m0.009s
> sys     0m0.024s
  Well, that's not really a useful number, since it's just doing a HUP 
and isn't checking results in any way.  I mean:

# timex kill -HUP 6631

real        0.03
user        0.01
sys         0.02

# time kill -HUP 6631

real    0m0.00s
user    0m0.00s
sys     0m0.00s

-Greg G