Package: diskless
Version: 0.3.18.0.5
Severity: grave
We've tried to setup diskless /diskless-image-simple and:
1) There seems to be an inconsistency in the naming scheme. Sometimes
configs are taken from the master from /etc/diskless-host (which does not
exist) and in some version from /etc/diskless-image... on the
client
only diskless-image is checked in /sbin/init , so booting a client fails.
2) If one fixes this one realizes that in /sbin/init only the /etc
filesystem gets mounted all the rest (/dev /var /tmp) the client is
supposed to deal with via the init.orig , i.e.
/etc/init.d/rcS -> /etc/rcS.d/S35mountall script. However that one does not
mount anything nfs related (nonfs nosmb etc in the very top of that file).
I suggest to mount everything in the /sbin/init wrapper like below (to not
having to modify the standard S35mountall script etc).
Furthermore the WARNING: master... is almost invisible as the system boots
to fast, so there should be a timeout like the sleep 5 inserted below
(as this is only for administration it won't delay the standard client
boot process). Also if the /etc/diskless-image/config does not exist or
the system cannot mount any /var /dev /tmp scripts for a client the system
should wait some minutes and then halt instead of go on booting...
---snip---
# Don't do anything special if root is rw (ie master system).
if touch /test-rw 2>/dev/null
then
rm /test-rw 2>/dev/null
echo "WARNING: master system installed" >&2
sleep 5
exec /sbin/init.orig $*
echo "FATAL ERROR: exec /sbin/init.orig failed" >&2
exit 1
fi
# Try and find configuration file
if [ -f /etc/diskless-image/config ]
then
# Load config file
. /etc/diskless-image/config
# Get IP address, need to mount /proc first
mount -n none /proc -t proc
IP=`ifconfig | grep -A1 eth0 | grep -v eth0 | sed 's/^.*inet
addr:\([0-9\.]\+\).*$/\1/'`
echo "Client IP address is $IP"
# Mount client's /etc directory
echo -n "Mounting NFS-root directories..."
echo -n "etc.."
mount -n $nfsserver:/$nfshostsdir/$IP/etc /etc -orw,nolock >/dev/null
# Update entries in /etc/mtab
rm -f /etc/mtab~ /etc/nologin >/dev/null
: > /etc/mtab >/dev/null
mount -o remount / >/dev/null
mount -o remount /etc >/dev/null
echo -n "proc.."
mount -o remount /proc >/dev/null
echo -n "var.."
mount $nfsserver:/$nfshostsdir/$IP/var /var -orw,nolock >/dev/null
echo -n "dev.."
mount $nfsserver:/$nfshostsdir/$IP/dev /dev -orw,nolock >/dev/null
echo -n "tmp.."
mount $nfsserver:/$nfshostsdir/$IP/tmp /tmp -orw,nolock >/dev/null
# report success
echo "done."
else
echo "ERROR: Cannot find image config file"
echo " Not mounting NFS-root directories"
fi
# Only should get here if something went wrong
exec /sbin/init.orig $*
echo "FATAL ERROR: exec /sbin/init.orig failed" >&2
exit 1
---snip---
3) On the master: The /usr/lib/diskless-image should clearly go to
/etc/diskless-image
and the former /etc/diskless-image should stay in /etc/diskless-host, as they
contain
config files (e.g. templates in /usr/lib/diskless-image/templates).
4) The /lib/modules directory is not client writable thus it should become a
symlink to say /var/lib/modules
4) Furthermore there seems to be no documentation at all. While there is
some xml.gz file in /usr/share/doc... no html/text documentation is generated,
as install-docs and dhelp are not found in the postinst script.
5) There are many minor issues like /etc/network/interfaces is getting
copied from the master (making the client fail as it gets the wrong network
settings) which could be fixed by supplying a
template/network/interfaces... Also the diskless-newhost update rules in
template/config require quite some updates to make certain programs work...
