Red Hat Bugzilla – Bug 76078
NFS install option ignores "discX" convention
Last modified: 2007-04-18 12:47:39 EDT
Description of Problem:
The RPMS merging is annoying for a network install. Much easier if one can just
loopback mount the CDs and NFS export them. Certainly easier if you have three
CDroms, or a jukebox for them.!
Version-Release number of selected component (if applicable):
Do an NFS install from a server setup like:
cassidy-[/distros/redhat/8.0] mcr 1161 %ls -lL
drwxr-xr-x 6 root root 6144 Sep 10 16:34 disc1/
drwxr-xr-x 3 root root 2048 Sep 10 16:34 disc2/
drwxr-xr-x 4 root root 2048 Sep 10 16:56 disc3/
-rw-r--r-- 1 mcr mcr 675315712 Oct 10 14:31 psyche-i386-disc1.iso
-rw-r--r-- 1 mcr mcr 666271744 Oct 9 16:46 psyche-i386-disc2.iso
-rw-r--r-- 1 mcr mcr 677609472 Oct 9 18:08 psyche-i386-disc3.iso
cassidy-[/distros/redhat/8.0] mcr 1163 %df | grep distros
662072 662072 0 100% /distros/redhat/7.2/disc
653740 653740 0 100% /distros/redhat/7.2/disc
650656 650656 0 100% /distros/redhat/8.0/disc
659488 659488 0 100% /distros/redhat/8.0/disc
661728 661728 0 100% /distros/redhat/8.0/disc
Steps to Reproduce:
It complains that /distros/redhat/8.0/ does not have a valid distribution.
I expect it to see that there are disc1/disc2/disc3 in that dir, and then do
three more NFS mounts for them.
(You have to do the NFS mounts, because NFS does not cross mount points on the
It would be nice if this worked for HTTP and FTP installs as well.
The workaround is to copy the RedHat/base/ dir from disc1 up one level,
and build a symlink tree from Redhat/RPMS/* to disc?/RPMS/*, and then do the
cassidy-[/distros/redhat/8.0] mcr 1216 %ls -l RedHat/RPMS/zsh-4.0.4-8.i386.rpm
lrwxrwxrwx 1 mcr mcr 44 Oct 16 12:32
RedHat/RPMS/zsh-4.0.4-8.i386.rpm -> ../../disc3/RedHat/RPMS/zsh-4.0.4-8.i386.rpm
cassidy-[/distros/redhat/8.0] mcr 1217 %cat domounts
mount cassidy:/distros/redhat/8.0/disc1 /mnt/source/disc1
mount cassidy:/distros/redhat/8.0/disc2 /mnt/source/disc2
mount cassidy:/distros/redhat/8.0/disc3 /mnt/source/disc3
For mirrors, it would be convenient if they could just suck the ISO images,
loopback mount them, and be done. That would halve their disk space.
That's why we support just using the ISOs remotely for NFS. Handling loopback
mounted ISOs on the server-side for NFS is much less convenient due to the way
NFS under Linux works.
I don't see how it matters. Three mounts, as I show, does the trick.
If there is a better way, I would love to hear it.
cf the documentation please. It is not obvious to me, and I've been doing
network installs of OSes since SunOS 3.