When I do clean install with a different partition scheme over existing RedHat 7.3 installation, anaconda sometimes failed to close the old disk fd. As the result, the new partition table won't be reread by kernel. I fixed the bug in RedHat 7.1 with Index: iw/rootpartition_gui.py =================================================================== RCS file: /work/cvs/redhat/anaconda/iw/rootpartition_gui.py,v retrieving revision 1.1 retrieving revision 1.2 diff -u -p -r1.1 -r1.2 --- iw/rootpartition_gui.py 10 Jun 2000 03:04:23 -0000 1.1 +++ iw/rootpartition_gui.py 11 Jun 2000 01:31:07 -0000 1.2 @@ -193,6 +193,10 @@ class AutoPartitionWindow(InstallWindow) clearcache = 0 # clearcache = not self.todo.fstab.getRunDruid() self.todo.fstab.setRunDruid(1) + # We have to do it. Otherwise, we may open the hard drives + # more than once. It may cause trouble when we revalidate + # the hard drives after we change their partition tables. + self.druid = None self.todo.fstab.setReadonly(0) #print "Rescanning partitions 2 - ", clearcache self.todo.fstab.rescanPartitions(clearcache) But this patch no longer applies to 7.3. I believe the bug is still there.
It happens with kickstart on an IDE HD connected to a Promise RAID controller.
More info. It doesn't happen with SCSI HD. After reboot, the installation works fine since kernel gets the new partion table.
What makes you think this bug is still there? Without verification that it's present in anything newer than 7.1 (and preferably, in GinGin), my inclincation would be to close current release
We no longer use Promise IDE RAID controller. I suspect it may be a bug in Promise IDE RAID driver. It only happened when a single HD was connected to Promise IDE RAID controller.
Time tracking values updated