Description of problem: I am seeing this on my HP Integrity servers with e1000 and e100 NICs (I do not have any other NIC's to test at this time). When doing a network based install and using DHCP to set up the networking anaconda gives the error "There was an error configuring your network interface." and it gives me a "Retry" button. I have seen this both with a kickstart install using ksdevice=eth0 as well as with an interactive install. I have also tried with and without the IPv6 box checked. When I do not use DHCP it works properly. Version-Release number of selected component (if applicable): rawhide-20060712 (note, this is the most recent good build for ia64 so I cannot test newer at this point). anaconda-11.1.0.56-1 How reproducible: 100% Steps to Reproduce: 1. boot the installer 2. select NFS, http, or ftp install source 3. select DHCP Actual results: Expected results: Additional info:
What messages are printed on VT3?
All of my systems are headless so I only have serial console. I really wish I had a way to get that info as I do realize it would make things much easier to debug. Is there some magic way to get at this?
You'll laugh, but you can use the remote logging feature. Send syslog messages to a remote system. Boot with these options: loglevel=debug syslog=somebox.location.redhat.com Set up syslogd to accept remote logging. *BUT*, since this is a network issue, that's not really going to help. You said that this is using anaconda 11.1.0.56. In that build, the version of libdhcp used still tried to write out /etc/resolv.conf and would fail. The anaconda loader was trapping that error successfully, but unfortunately it wouldn't let you proceed and you'd end up in the retry loop. If you can get a tree that's using anaconda-11.1.0.57, that's built against a libdhcp that does not have that problem.
Dave, This doesn't help me right now but the is AWSOME for other anaconda issues. I will certainly try the remote logging next time I see an anaconda issue! I cannot currently use more recent trees because of panic and build issues but I will re-try this as soon as something is available. thanks, - Doug
Doug, Sounds good. The retry loop is most likely the /etc/resolv.conf writeout, which has been fixed. Also, I'm going to close 196701 since it's sort of the same bug as listed here, but this one is in a more advanced state. -David
Doug, Where are we on this bug? Is it still present in current rawhide trees? -David
David, sorry, should have closed this one out a while back. Seems to be working now.