Bug 74910
Summary: | Installer shouldn't use dhcp hostname by default. | ||
---|---|---|---|
Product: | [Fedora] Fedora | Reporter: | Alan Cox <alan> |
Component: | anaconda | Assignee: | Anaconda Maintenance Team <anaconda-maint-list> |
Status: | CLOSED CURRENTRELEASE | QA Contact: | Mike McLean <mikem> |
Severity: | medium | Docs Contact: | |
Priority: | low | ||
Version: | rawhide | CC: | julo42, leenookx, lee.wilson, notting, otaylor, sopwith |
Target Milestone: | --- | Keywords: | FutureFeature, Triaged |
Target Release: | --- | ||
Hardware: | i386 | ||
OS: | Linux | ||
Whiteboard: | |||
Fixed In Version: | Doc Type: | Enhancement | |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2005-06-07 19:33:45 UTC | Type: | --- |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: | |||
Bug Depends On: | |||
Bug Blocks: | 150221 |
Description
Alan Cox
2002-10-02 18:49:06 UTC
I don't personally understand this issue well enough to know what we need to offer to fix (or if it can be autofixed). If it can be autofixed we could maybe just do so without asking (or just get it right in the first place). The right fix depends on the factors beyond the local host. Sometimes the 'right fix' lies in the DNS database on some other server (to make matching forward & reverse entries), while other times it can involve adding the FQDN and/or hostname onto the 127.0.0.1 line in /etc/hosts. I can't remember the third case, but deciding between the three cases requires a human. Having a button to launch network configuration would be a good thing but would need the window manager running, which it isn't. Could we run the window manager then kick it off, or run it under the mini window mangler ? I guess we could check to see if /etc/hosts corforms to the Red Hat standard, and if it doesn't do it, run a user-helpered tool to fix it. I'm not sure what the Red Hat standard is currently, though. What was fix you ended up doing, Alan? I added the names to the 127.0.0.1 entry. That had old names from before I set the hostname. Once that was done all was fine. It is a user intervention item in many cases because its just as possible you want to know and set up the missing DNS records that way I have had this happen on my box when I select DHCP during a fresh "Personal Desktop" installation. When DHCP is selected as the network card's means of get an IP you are not prompted for a hostname, consequently none gets added during the install. Could it not be made to just use a default hostname, something like "redhat.localdomain". I believe during debian installation it uses this as a default if you don't want to choose it yourself. This problem should be easily enough fixed (as someone else has mentioned above) by setting the hostname correctly in "/etc/sysconfig/network" and also adding the appropriate entry to "/etc/hosts". My first impressions of RH8 is that it is more designed for the new Linux user so in a similair vain we should be looking for a way for this to be resolved automatically rather the relying on the user to user it out. Apologies if I have just waffled on about something completely wrong. I'm all for fixing this longstanding issue but my knowledge of network configuration is near zero, so I think we should reassign the bug to someone with a clue. I'll write GUI for it if someone clearly specifies what the GUI should do. *** Bug 69315 has been marked as a duplicate of this bug. *** I just want to talk about my personal experience: I have a laptop connected to an ADSL ethernet modem. I set up my internet connection, during the install, so that it uses DHCP to determine the IP address and the hostname. When I first log in using GDM, GNOME (which is launched by GDM) warns me that my hostname can't be found (or is incorrect...) and tells me to add "dhcppc0" (the hostname given by my ADSL modem) to /etc/hosts. I added it to /etc/hosts (on the 127.0.0.1 line) and all was OK then. So my guess is that, for most users, a simple solution is to automatically add the hostname to /etc/fstab when the eth0 connection is started. That's only my end-user point of view though. I don't know which problems that could imply. What I know, however, is that it works automatically on Windows XP. So there must be a solution to this problem. Hi, I'm going through bugs assigned to me and attempting to clean some of the older, fixed ones up. This bug hasn't changed in over a year old now. Are you still seeing the problem? (This a batch message is being sent to all my bugs that haven't changed in a year) Hi, So I talked to Elliot about this a couple of weeks ago. We do need a hostname that maps to an ip address for two cases: ORBit and X auth cookies. The hostname given doesn't have to be externally resolvable or have any significance outside of the machine it's working on. The name does get shown to the user in various places, so using whatever name a random ISP's dhcp server gives is a bad idea. Ideally when the user installs Fedora, the installer would ask the user "What name do you want to give to your computer?" or something and then would normalize the input into a legal hostname. An entry for this could automatically be added to /etc/hosts and all programs that depend on hostnames that are forward and reverse resolvable should have no more problems. For weird cases like lab setups where it still might be useful to get the hostname from dhcp, the admin can always just change /etc/sysconfig/network or whatever. I think we actually handle this properly now. The installer asks which you want to do and defaults to DHCP if you're using DHCP. But the initscripts also do the right thing of falling back to localhost as your hostname if you don't get a hostname via DHCP and asked for that. |