On Tue, 8 Feb 2005, Martin Hamant wrote: > I install Centos'es from PXE, with vmlinuz and initrd provided by > images/pxeboot from the first CD of centos 3.4. > > The server is a PowerEdge 1850 with two integrated e1000 nic interfaces. > > So kernel is booting, then it comes to the screen which ask for one nic > to do a dhcp query. I choose eth0, dhcp server reply well and then i get > an ip address, dns'es, etc. And nothing else happen because the nic link > is going down at this point !!! (no light, no way. ) There have been changes over time of the MII link management code used by Anaconda. I encounter these issues with 3-com nic's and Intel switches, but have also encountered them with T-Lan nic's as well. Simplify the network topology - go to a straight crossover to your dhcp/next server if possible - alternatively, consider trying a single switch between the unit under test and the dhcp/next server -- then vary the switch brand, to test each. If possible, run TCPDUMP on the server side, and see whenre the connection is failing. -- Russ Herrold