dhcpd configuration

DJ Lucas dj_me at swbell.net
Mon Apr 22 16:54:12 PDT 2002


Larry, 

I'm thinking additional notes tacked on to the end of the chapter as an
FAQ would be the perfect place for noting the change in the kill scripts
and notifying the user that if they upgrade bootscripts, they'd need to
remove K80network as well.  That seems like a great way to handle it.
There is one more thing that I've noticed about dhcpcd that really is
not an issue for most networks but also IMHO, worth mentioning in the
FAQ if it is created.  Killing the PID for the dhcpcd client does not
send a DHCP release to the DHCP server, nor does ifconfig down.  This
will result in a valid lease being reserved until the lease expires.
The only place I can see where this would be an issue is when a DHCP
server is set up with a small address pool; Less than the number of
total clients, but not less than the number of simultaneously connected
clients.  This would typically be used in larger departments of a
corporation to reduce the amount of addresses needed for a particular
network segment..basically for a shift change (3 shifts) in my following
example.  I have actually seen such a network, and this very problem
played terribly on the poor admin I was calling at 3 in the morning
during a PC deployment. Not to mention the fact that my techs,
unbeknownst to me, were a partial cause of the problem. :(  Anyway, this
may not be the only situation where this could be a problem, but it's a
familiar one to me.  Again, I'm writing this as I found it, and haven't
taken the time to explore how dhclient handle's this situation.  I'd
assume the same, but, I know quite well what assumptions lead to. ;-)  I
do however know that killing the clients directly with their respective
switches will send the correct release.  I'll update on dhclient later
this evening.

Thanks again,

DJ Lucas

________________________________________________________________________
____
Larry Lawrence wrote:

> DJ Lucas wrote:
> 
> > I don't know what I was thinking nor why I thought you should change
> > that.  One more question.. is it necessary to use killproc in the
> > script?  Doesn't sendsignals take care of that for us anyway?
Assuming
> > the user does modify the kill scripts, then this will cause an error
> > with multiple NICs using the same client.
> > 
> > Thanks for all the help here,
> > 
> > DJ Lucas
> > 
> I think a lot of users believe the killing the dhcp client brings down
the 
> interface. I did not give it any thought until testing the scripts.
The 
> test I ultimately used to finalize the scripts was to drop to runlevel
2 
> from 3.  By LFS definition, the network must be down. There is no kill

> sendsignals in this level (it would be too destructive).  I believe I
also 
> found the ./network stop failed to clear the pid on dhcpcd.  Multiple
dhcp 
> NICS is an issue that I need to look at.
> 
> Moving K90network to K49network is my mailing list solution and will 
> probably be a FAQ for eliminating the warning, my problem with putting
it 
> in the book is the question of how much warning.  For example, do we
need 
> to let them know that the next time they upgrade bootscripts, they
will 
> have two Kxxnetwork's. 
> 
> Glad you subscribed to the list.  Keep those comments and suggestions 
> comimg.
> 
> Larry  

-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe blfs-dev' in the subject header of the message



More information about the blfs-dev mailing list