Red Hat Bugzilla – Bug 64599
Bad error reporting in anaconda
Last modified: 2007-04-18 12:42:28 EDT
Description of Problem:
Installing redhat 7.3. Get error message as per
http://student.cs.ucc.ie/~cjvdb1/images/rh73inst/p5070292.jpg saying disk is
probably full. According to df in
http://student.cs.ucc.ie/~cjvdb1/images/rh73inst/p5070292.jpg disk has many
gigabytes still free.
This mystified me at first, but I have since discovered that the real error
would appear to be a read error on the cd (another copy of the CD installs
Version-Release number of selected component (if applicable):
Don't know. The one on valhalla-i386-disc1.iso :)
Every time, with particular CD.
Steps to Reproduce:
1. Boot from CD containing package with read error (in this case,
2. Proceed with upgrade install.
3. Install falls over when it gets to unreadable package.
Install fails because of one unreadable package, and gives inappropriate error
message (disk probably full).
Successful install, skipping unreadable package.
See http://student.cs.ucc.ie/~cjvdb1/images/rh73inst/anacdump.txt and images
in that directory.
Another traceback for a similar problem (no disk full error message, just a
crash with traceback) is available from
i checked all my cds - 1 said it passed the test - the other two failed. all
iso images matched the check sum. most i can figure is that it is a bad burn.
i used xcdroast from the beta 2 version. this is bug 64599 seems to be a dupe
workaround i copied all the files to one dir and did a http install, worked fine.
It is recommended you use the mediacheck option at boot time to catch the CD
It's an imperfect world; coasters exist. I would imagine read errors are one
of the most common errors during an install - why not report them properly?
"You are probably out of disk space" is very misleading. Other OS installers
give various options in this situation (abort/retry/fail ring any bells?); why
does RedHat only offer abort/debug? (where debug also leads to abort)
MY CDs are fine. The problem seems to be that the openjade rpm package causes
the failure. The upgrade installation expects the openjade rpm on CD#1, but it
is packaged on CD#2. When the package is not found, the installer aborts. If all
rpms are in the same directory (disk, ftp, or http install) or openjade is not
one of the packages to be installed, the problem does not exist. See the
anacdump.txt files in the URLs above.
I concur with cjvdb1 -- the problem here isn't that the install crashes, it's
the Microsoft-quality error message that's given, and the fact that the error
I propose to file a new bug with this adjusted description, if no one sees fit
to reopen one of the currently extant ones, and I will *not* tolerate a NOTABUG