[rhelv6-list] dhclient running TWICE during boot (RHEL 6.2)

Paul Smith paul at mad-scientist.net
Mon Apr 30 14:29:17 UTC 2012


On Mon, 2012-04-30 at 10:04 -0400, Bill Nottingham wrote:
> Paul Smith (paul at mad-scientist.net) said: 
> > My problem is the interface is controlled by both and I want it
> > controlled by NM.  How do I tell the network service to NOT try to
> > control it?  Is setting NM_CONTROLLED=yes sufficient to tell the network
> > service to ignore that interface?  If so why doesn't it get set by NM
> > when it starts to manage the interface?  I've looked at the scripting
> > for the "network" service and I don't see where setting that parameter
> > will help.
> 
> The network script will use NetworkManager if it's running and redirect
> ifup to a call to NM to start the interface. However, if NM isn't running
> yet, it will bring the interface up itself.

During boot, when S10network is run, NM is not running yet...

> That being said, NM *should* just take over the valid lease when it starts;
> if it doesn't, something else is going on.

Not sure what you mean, technically, by "take over the valid lease".  Do
you mean, when NM starts its dhclient instance it gets back the same
results from the DHCP server (since it sent the request with the same
MAC), and so both invocations of dhclient have the same result?  I do
notice only one dhclient is running once the system is up, and ISTR
seeing /var/log/messages entries to the effect that once was being
killed, so that part is working, but the leases are definitely NOT the
same.

I don't know why the DHCP server is providing different IP addresses but
this is a very widespread phenomena on my network; all the RHEL 6.2
systems I've looked at, even ones where the IP address and DNS entries
match, have both these lease files and the IP addresses are different.
I believe that our IT department is using some Windows server based DHCP
service; maybe it behaves differently than other DHCP servers and if you
request a new lease before the previous one was ACK'd it hands out a new
IP address, or something.

I guess I can try running tcpdump during boot to try to figure out
what's going on there.




More information about the rhelv6-list mailing list