Red Hat Bugzilla – Bug 138140
Discovery does not create valid HARDWARE.conf
Last modified: 2007-04-18 13:14:38 EDT
Description of problem:
After installing Redhat Ready in both AS V3 U2 and ES V3 U2, I have
discovered that the Discovery phase of the program is not consistent
in creating a valid HARDWARE.conf file, in the /etc/rhr directory.
This problem results in testing that fails, or testing that passes
due to no parameters being passed to the tests
in /usr/share/rhr/test. Most notably, I have reproduced, three
times, a HARDWARE.conf file with nothing in it but the "disclaimer".
In running the NETWORK tests, with no ethX entries to pass, all those
tests fail. Furhter, in one instance where the HARDWARE.conf file
was populated, an erronous "N/A" was added to the descriptor line for
the STORAGE tests. This resulted in the storage tests failing
the "badblocks" test, as one of the passed /dev/$device arguments
I was able to debug to the HARDWARE.conf file "un-population" as the
culprit by stepping through the tests in the /usr/share/rhr/tests
directory, and seeing that the tests were either being passed nothing
as arguments (in the NETWORK case), or they were being passed invalid
arguments (as in the STORAGE case).
Version-Release number of selected component (if applicable):
Most of the time (sometimes the HARDWARE.conf file is empty,
sometimes it has bad data in it)
Steps to Reproduce:
1.Install AS, or ES V3 Update 2, with Everything
2.Install Redhat Ready, prep machine for testing (gathering files,
CDs, blank floppies, etc.)
3.After pre-execution services started (Apache, NFS), start RHR, run
Discovery, and check HARDWARE.conf in the /etc/rhr directory for
completeness. No third-part RPMs are installed in this test scenario.
As described above, the HARDWARE.conf file in the /etc/rhr directory
is populated with nothing, or has bad data for devices under test.
The HARDWARE.conf file should actually contain the correct number of
devices found; or if the discovery doesn't work, inform the user that
issues were found in creating the file.
In trying to debug both failures (in NETWORK and STORAGE), I found
the lack of reporting capabilities in to the cooresponding error.log
and output.log files to be agrivating. Given this is a set of tests
for use by technicians who are trying to "certify" thier HW with the
RH products, I would like to see more debug style statments put into
the test files and scripts. For instance, the STORAGE tests could
easily have the $device, $beginning_block, and $ending_block entered
into one of the above mentioned files, providing the tester valuable
information for troubleshooting.
Problem is hardware.py (used by discovery to describe the
hardware/provided by up2date) requires connection to
xmlrpc.rhn.redhat.com; satisfy this http request (doesn't need
specific info; simply an http server responding will do).
A work around has been committed in a later release (beta available on
http://people.redhat.com/rlandry/rhr2/updates/ [NOT FOR USE ON REAL
HWCERTS] ). In addition, a sample hardware.conf has been included in
the docs as discovery is to make things easier not required, manual
generation of hardware.conf is acceptable (and possibly required) in
cases where discovery does not detect everything.
In addition, "-d" is available for debug logging (though it has the
side effect of causing all tests to report failed).
*** This bug has been marked as a duplicate of 118154 ***
Changed to 'CLOSED' state since 'RESOLVED' has been deprecated.