Red Hat Bugzilla – Bug 88023
anaconda crashes during network install with kickstart: TypeError: bad magic in header
Last modified: 2007-04-18 12:52:46 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020830
Description of problem:
I generated an 8.0 kickstart file using the 8.0 Kickstart Configurator 2.3.3. I
modified the resulting kickstart file adding certain RPM groups and added a
%post section. I put the kickstart file on a network boot floppy, and booted
with ks=floppy (ks file attached).
The machine boots fine, it uses method http. Monitor is auto-probed fine. It
retrieves netstg1.img successfully etc from boot server. It loads the megaraid
driver (onboard raid) and the AIC scsi card driver and then anaconda crashes
producing a tracedump (file attached). The last entry in the httpd log of the
x.x.x.x - - [04/Apr/2003:13:40:08 -0500] "GET
/data/iso/mnt/redhat/8.0/RedHat/base/hdlist HTTP/1.0" 404 1044 "-"
Since I am doing iso loopback mounts, it should check
<installbase>/RedHat/base/hdlist *and* <installbase>/disc1/RedHat/base/hdlist,
but it never does. Anaconda crashes before it checks the second location.
The bootserver is running RedHat 8.0. It serves out installs via http, using
iso9660 loopback mounts. I can verify that the http distribution works as I can
wget files from the install url from other machines.
The computer running the bootfloppy is a dell poweredge 4200, with scsi drives.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Create an 8.0 network boot floppy using dd.
2. Remove snake.msg to make room for the ks.cfg file.
3. Modify the syslinux.cfg file to boot with default ks=floppy (attached).
4. Create a ks file with the Kickstart Configurator 2.3.3.
5. Modify the ks file to add custom RPM's
6. Modify the ks file to add custom partitions
7. Modify the ks file to add a %post section.
8. Boot machine with floppy.
9. The crash.
Actual Results: Recieved an "Exception Occurred" window with an anaconda traceback.
Saved the crash dump to floppy.
Expected Results: Should have gone on to automatically configure partitions.
Relevant files attached, and present at URL above.
Created attachment 90903 [details]
Anaconda trace dump.
Created attachment 90904 [details]
HTTP access log showing that anaconda didn't check for disc1/RedHat/base/hdlist
Created attachment 90905 [details]
My (slightly modifed) kickstart file.
Created attachment 90906 [details]
My (slightly modified) syslinux.cfg file that I use on my network boot disk.
What does the hierachy under:
I was able to do an installation using loopback mounted ISO images w/o problems.
Here is the hierarchy:
[esm@sys02 8.0]$ pwd
[esm@sys02 8.0]$ ls
disc1 disc2 disc3
Did you use my kickstart file to do the install? The question isn't whether the
iso/loopback mount will succeed - I've performed that sucessfully in the past.
I don't think anaconda should be failing....
FYI you should be able to browse the hierarchy for yourself:
http://vulcan.mse.jhu.edu/data/iso/mnt/redhat/8.0 (added the fqdn to the url)
If you add '/disc1' to your kickstart install url spec does that help?
Admitedly I tested this with RHL 9, in which I seem to recall I may have fixed
an issue like this...
No this did not fix the problem.
Have you looked at the attachements I've provided? If you look at the http
access log you will see that there isn't an issue with finding netstg1.img nor
updates.img both which sit under disc1. The installer doesn't know if I have
the redhat installation under <installbase> or if I have it under
<installbase>/disc1 so it tries both - that is built into the installer. The
install performs fine if I don't use kickstart with the same install base of
You also didn't answer my first question: did you try the install with *my*
kickstart file and have you looked at the trace dump at all?
Created attachment 90920 [details]
Patch to urlinstall.py
This is a patch to /usr/lib/anaconda/urlinstall.py in diff -u format. It
compiled successfully but I have yet to remake the netstg1.img file. Will test
Ok, I feel stupid. I've been using /data/iso/mnt/redhat/8.0 as my base url,
instead of /data/iso/mnt/redhat/8.0/disc1. I didn't think I had to specify
/disc1, I thought it would auto-detect whether or not i was doing a cd
installation (based on the behavior of the installer when it tried retrieving
netstg1.img and update.img). So the previous patch is bogus. Less coffee and
more sleep would do me some good I think. Thanks for your help!
> If you add '/disc1' to your kickstart install url spec does that help?
As an aside, I misunderstood what you meant by this comment, but now I do!