I have the same troubles with CentOS 6.4 LTSP client and others.
Fortunately I have an older client image based on Scientific Linux 6.1
which works fine. I tried to compare both of them. It is difficult for
me, and I didn't solve troubles but let me share my notices.
CentOS 6.4 client image has empty /etc/mtab file. It's true, but if the
client is booted you can check this file on a client on local xterm -
this file is not empty.
Scientific Linux 6.1 client image has symbolic link between /proc/mounts
So, I suppose that LTSP 5.4 uses different mechanizm.
closing the client
When I try to close CentOS 6.4 client image I can see that /tmp/sysroot
and /tmp/unionfs/rofs directories are unmounted and closing system
suspends on shutting down eth0.
Trucking Scientific Linux 6.1 client image I made as chroot /sys
-> /tmp/sysroot symbolic link. Now closing CentOS 6.4 client I don't see
message about unmounted /tmp/sysroot directory. So, I suppose it helped.
But still I can see message that /tmp/unionfs/rofs is unmounted and the
client doesn't power off.
Let me add two other points.
The first one:
If I try to add some package to CentOS 6.4 client image I use:
# setarch i386 chroot /opt/ltsp/i386
# mount /proc
But I can see "can't find /proc in /etc/fstab or /etc/mtab" message.
I can install additional packages, but sometimes I can see message
that /proc is empty. If I add to /etc/fstab "proc /proc proc defaults
0 0" line all work fine.
What interesting, if /etc/fstab is empty and I check this file on booted
client - fstab is not empty.
So, I don't know - does it makes sens to add to /etc/fstab "proc /proc
proc defaults 0 0" line, or not? I added.
The second one:
Still CentOS 6.4 client. When LDM is loaded and I try the first time to
log in the greater reloads. Te second log in is effective. I expect that
the first trial should be effective.
There are the reasons that I use Scientific Linux 6.1 client image.
Post by Peter Onion
I've build clients on a Centos 6.4 platform and after some "tweaks" they are working, but they won't shut down properly.
This is cause by the "stop" sections in network,nfs and nstfs scritps in /etc/init.d all failing to identify that the root file system is mounted over the network.
stopping eth0, nfs or trying to umnount everything has predictable results in this case !
And the halt script has a go as well when it kills the unionfs user space process and then everything stops !
Also "out of the box" /etc/mtab ends up as an empty file which I don't think is correct.
I've got the clients to shutdown or reboot properly by adding "exit 1" at the beginning of the "stop" sections in the abve mentioned scripts.
Is there a better/correct solution to this ?
K12OSN mailing list
For more info see <http://www.k12os.org>