Bug 499408

Fwiw, I pulled a couple older releases to test.  Fedora 11 Beta fails with the same problem.  Fedora 11 Alpha works.

and Rawhide install via netboot fails the same way

Still an issue on latest rawhide with anaconda-11.5.0.51

I will do some more digging on this.

Still digging into this and learing more about how to debug anaconda.

I have found that it works OK when there are not any existing partitions on the cciss disks.  The traceback comes from when it tries to get info on existing partitions using getDeviceByName.  Looking at what data is available at that point I see the disk entries (i.e. cciss/c0d0) but I don't see that it ever added entries for partitions (such as cciss/c0d0p1)

(In reply to comment #4)
> Still digging into this and learing more about how to debug anaconda.
> 
> I have found that it works OK when there are not any existing partitions on the
> cciss disks.  The traceback comes from when it tries to get info on existing
> partitions using getDeviceByName.  Looking at what data is available at that
> point I see the disk entries (i.e. cciss/c0d0) but I don't see that it ever
> added entries for partitions (such as cciss/c0d0p1)  

Where does it get the devicename from it passes to getDeviceByName ? If it is from a device node name, such as for example /dev/cciss/c0d0p1, the correct
way to get the name from the device node path is to use devicePathToName(),
some places used to do devicepath.split('/')[-1], which is wrong, maybe I missed one.

Created attachment 344054 [details]
patch to fix install on cciss

since cciss device names have a subdirectory (i.e. cciss/c0d0p1) they are represented in sysfs with a ! so that it doesn't incorrectly have an extra subdir in the sysfs namespace.  When anaconda gets the device name from sysfs it was trying to lookup the name with a ! but anaconda stores them with the / in the name.

Simple 1 line fix is attached.  Please take this patch for F11 as nearly the entire HP Proliant line will be broken without it (nearly all of these systems use cciss for storage).

I second that request.  Otherwise F11 will fail to install on practically all 
ProLiant servers and we'd really like to get some testing with 
Fedora before there are RHEL6 alphas. 

BTW, we also noticed in the F11 alpha, which does install on a ProLiant,
that the cciss driver was missing from the initrd.   It was reported
in https://bugzilla.redhat.com/show_bug.cgi?id=484260 an appeared to have
been fixed but it seems to be a chronic problem so hopefully it is 
still fixed.

Doug, thanks1

The patch has been applied to the current anaconda tree, and will be
in the next build of anaconda (11.5.0.53), can you please test a rawhide
install when 11.5.0.53 is in rawhide, and report if cciss installs work now ?

I don't see this in rawhide yet. Will it be in F11? It would be nice if we could test a couple installs with this patch before release.

(In reply to comment #9)
> I don't see this in rawhide yet. Will it be in F11? It would be nice if we
> could test a couple installs with this patch before release.  

The fix is in this build:
http://koji.fedoraproject.org/koji/buildinfo?buildID=102607

Which has an f11-final tag, so it should show up in rawhide rsn.

I see this on the internal redhat fedora mirror so it just has not made it out to all mirrors yet.  I did a quick test and it appears to be working properly on cciss now.  I will do a couple more installs with different cases but it looks like this should be resolved.

This bug appears to have been reported against 'rawhide' during the Fedora 11 development cycle.
Changing version to '11'.

More information and reason for this action is here:
http://fedoraproject.org/wiki/BugZappers/HouseKeeping