Duncansapien’s Techblog

Overview

If we have a working Virtual Machine, all setup nicely and having had the disk resized perhaps, and partitions enlarged, it may be easier to use this as the basis for a new VM for another user. The cloned machine requires to be configured as follows.

Steps.

1. Physically copy the VM files to a new folder.
2. The VM filenames may wish to be changed to reflect a change in hostname, for example. Open the main .vmx VMware configuration file and change the names of files inside from e.g. suse-somac8 to suse01, whilst at the same time renaming the config files referred to in Explorer.
3. There may be a .vmxf extended configuration file also, for which the above process should be performed.
4. Start the VM and login.
5. Su to root and
   • change the root password
   • delete any unwanted user accounts
   • Launch Yast
6. In Yast, we require to rename the VM and set the correct network information.
   • This (great) page tells you how to rename the Linux server and change the IP information: http://www.swerdna.net.au/linhowtonic.html
7. Also one must manually set the IP number and default route or the network card does not work (Bridged networking - IP number on SML subnet, not DHCP etc)
   > ifconfig eth2 137.195.80.82 netmask 255.255.255.0 up
   > route add default gw137.195.80.90
   where the above 80.82 respresents the guest IP number and 80.90 is our default gateway.

Installing VMware Tools

Vmware tools are basically drivers that work well in the Virtual environment - e.g. graphics drivers. One chooses to install them from the Vmware software and this appears as a Virtual CD (which can be mounted) and on which the vmware tools RPM is found.

There is a gotcha. Vmware Tools requires the kernel source code - just a folder which contains the source code. On the version of Suse in use here, the running kernel and source-code folder do not match for some reason, and VMware tools does not install. The steps to remedy this are

1. Run command uname -r to return the running kernel version 2.6.16.21-0.13-default
2. Yast - Software manager - search for kernel-source and examine the versions. Version 2.6.16.21-025 is installed, hmm.
3. From this RPM repository, search for the correct kernel-source and download it (to a USB stick, say). Actually I found this kernel here: http://mirror.math.temple.edu/suse-install/update-10.1/rpm/i586/
4. Install the above kernel-source

The kernel-source now matches the running kernel (you may have to remove the existing kernel-source within Yast, amidst many warning and protests!)

1. Now with the Vmware Tools virtual CD, copy the vmware-tools.x.tar.gz file to and install folder of your choice (where x represents a lot of numbers I’m too lazy to type).
2. Run command
   > gunzip vmware-tools.x.tar.gz
   > tar -xvf vmware-tools.x.tar
3. Change to the newly created vmware-tools-distrib folder
4. Type ./vmware-install.pl to run the installation script.
5. Accept the defaults and when asked if you want to run a config script, accept. Here’s where a problem comes, because at some stage a default answer will not suffice. You’ll get an error message if you accept the default to question "what is the location of the C header files that match your running kernel [/usr/src/linux/include]?"
   You require to locate a folder such as /usr/src/linux-2.6.16.21-0.13 which should match your running kernel version, or the script will loop here. This is why the kernel-source was corrected earlier

Post Installation

It was found that the screen resolution was not able to be changed, even after a restart.

1. Enter single-user mode
   > su
   > init 1
2. Login as root and start the graphics configuration routine:
   > sax2
3. The X-windows environment should start. Under Monitor, change the monitor’s resolution to what matches your system (here 1280×1024).

Checking non-root partitions

To be added.

Checking the root partition

The root partition should not be mounted in order to check it, which means we must

Reference:

1. Invalid Superblock Error:
   • "If mounting a file system fails due to an invalid
     superblock, the e2fsck program would probably
     fail, too. If this were the case, your superblock may be corrupted, too.
     There are copies of the superblock located every 8192 blocks (8193, 16385,
     etc.). If your superblock is corrupted, try one of the copies instead. This
     is accomplished by entering the command e2fsck -f -b 8193
     /dev/damaged_partition. The -f option forces
     the file system check and overrides e2fsck’s
     error so that, since the superblock copy is intact, everything is fine." (Reference)
2. The dd command - http://www.debianhelp.co.uk/ddcommand.htm
   • Backup the MBR partition with syntax: dd if=/dev/hdx of=/path/to/image count=1 bs=512
   • Example: dd if=/dev/hda of=backup-of-hda-mbr count=1 bs=512 (from this Ref).
   • The actual command used were:
   • > fdisk -l [to check partitions, reports sda1 as swap and sda2 as root /]
   • > dd if=/dev/sda1 of=backup_of_sda1_mbr count=1 bs=512 [in root folder]

Making Apache case-insensitive:

Scenario:

Html
files are held on the Scratch drive of a Novell NetWare 6.0 server, currently
served by a Apache32 which is case-insensitive. A Linux solution is
naturally case sensitive, so we need a way around this if we are to replace the Windows Apache server with a Linux one.

Versions:

• Novell NetWare 6.0
• Suse Linux 10.1 with Apache 2.2.0-21.7

Links:

• Mount webroot as a Samba share
• Use module mod_speling

Here’s the Plan

1. Mount the Novell server volume holding the Apache htdocs folder using ncpmount
2. Create a Virtual Host on Apache which will use this htdocs as it’s document root
3. Use mod_speling in this vhost to ignore case.

And here’s what actually worked

1. Mount the Novell server volume holding the Apache htdocs folder using ncpmount
2. When
   creating a vhost to serve the Novell htdocs folder, Apache will not,
   for some reason, serve html documents. It will serve text files, but
   not .htm or .html files and nothing appears in the error logs. Quite
   bizarre.
   • As a workaround, create a normal directory for
     the Vhost in question and use rsync to copy the Novell directory
     contents into this vhost directory. Long file names are copied just
     fine.
3. Load the mod_speling module into the Vhost config file and indeed, URLs are now case insensitive.

Important: DNS and Name Resolution in this article

In using the Vhost name of sml2.som.hw.ac.uk it’s important to remember that, in this test, this name is not actually DNS registered. To get the Linux machine to resolve this one must add this to the /etc/hosts file, with the same IP number as the suse01.som.hw.ac.uk host itself. On Windows, one must edit the hosts file under C:WINDOWSsystem32driversetc by including the line:
137.195.80.104 sml2.som.hw.ac.uk itc.som.hw.ac.uk

The Files Involved in this article:

a) The Novell Server mount point

• > mkdir /mnt/somfs2 [as root]
• > chown wwwrun.nwaccess /mnt/somfs2 [so that the Apache process - wwwrun- can access the folder]
• > chmod a+rx /mnt/somfs2 [so all Linux users can read this mount]
• > cd /mnt/somfs2
• > chown wwwrun.nwaccess . [so that Apache owns the folders/files actually mounted here]
• > chmod a+rx . [so all Linux users can read this mount, makes things easier]

The SCRATCH volume of Novell server SOMFS2 is mounted via the command

• ncpmount
  -S somfs2 -A somfs2.som.hw.ac.uk -V SCRATCH -U somweb.som.staff.hw -P
  thepassword -u 30 -g 1000 /mnt/somfs2 -d 0755 -f 0755

Here, somweb is the Novell webserver login, whose password is given after -P. The -u and -g
options specify that the mount process is to be owned by the user with
uid 30 (wwwrun) and group having guid 1000 (nwaccess); both can be
determined by viewing /etc/passwd /etc/group.

The -d and -f options
specify the file permissions with which the folder is to be mounted;
the first digit - a zero - indicates … I’ve forgotten, but 755 is
u=rwx,g=rx,o=rx.

In testing, the above command was
saved as a Bash shell script in /home/somdjp/apache_mount.sh and called
by root but ncpmount is flagged suid. Thus, the command runs with
privilieges of the file ownership, which is root, so it need not be
called by root. Check this with ‘>whereis ncpmount’ and ‘ls -l
/usr/bin/ncpmount’ which will return -rwsr-x—.

After mounting, we have this directory structure (where the red text represents folders on the Novell folder): /mnt/somfs2/Apache/htdocs/

b) The Apache Virtual Host: Config File

A virtual host (Vhost) was created by means of creating file /etc/apache2/vhosts.d/vhost-sml2.som.hw.ac.uk.conf

The contents of this file are below, the most important parts being documentroot and mod_speling.

<VirtualHost 137.195.80.104>
ServerAdmin D.Potter@hw.ac.uk
ServerName sml2.som.hw.ac.uk

# Duncan: Enable mod_speling

LoadModule speling_module /usr/lib/apache2/mod_speling.so
CheckSpelling On

DocumentRoot /srv/www/vhosts/sml2.som.hw.ac.uk/htdocs

# Default MIME typer - for text and html documents
DefaultType text/plain
Loglevel error
Alias /icons/ /srv/www/vhosts/sml2.som.hw.ac.uk/Apache/icons/
</VirtualHost>

c) The Apache Virtual Host Document Root

Remember, this is a real directory whose contents will be served by the Apache vhost. The contents of this folder are the same as the Novell mounted htdocs folder, and the rsync program will be used to copy the contents across. For some unknown reason, despite much effort, the Apache vhost cannot serve html documents from the Novell mounted htdocs folder, so we have to use this mechanism to get past this.

The documentroot is /srw/www/vhosts/sml2.som.hw.ac.uk/htdocs.

Using Rsync

In testing, only one folder under the Novell mounted htdocs was copied, and this was gb6.

• Novell: \somfs2scratchapachehtdocsgb6
• Suse: ncpmount’ed as /mnt/somfs2/Apache/htdocs/gb6
• Suse: rsync’ed to /srv/www/vhosts/sml2.som.hw.ac.uk/htdocs/gb6

The rsync command was [rsync options source destination]

• > rsync -avr /mnt/somfs2/Apache/htdocs/gb6 /srv/www/vhosts/sml2.som.hw.ac.uk/htdocs/gb6

which duly populated our Vhost htdocs documentroot. To remove files in the destination which are no longer in the source one would use the –delete option:

• > rsync -avr –delete /mnt/somfs2/Apache/htdocs/gb6/ /srv/www/vhosts/sml2.som.hw.ac.uk/htdocs/gb6/

Making Apache Case Insensitive

The subject of this article! All that’s required is

1. Track down where the Apache modules are held
2. Add the following lines to the Vhost.conf file in question (vhost-sml2….conf):
   • LoadModule speling_module /usr/lib/apache2/mod_speling.so
     CheckSpelling On
3. Restart Apache
   • > rcapache2 restart

Apache is now case-insensitive. Actually, when you type in a URL using the wrong case, Apache finds the file anyway and serves that so the web-browser location shows the proper, Linux filename. However, the case-insensitivity only applies to file names - not directory names. Example

• Correct URL is: http://sml2.som.hw.ac.uk/gb6/BroughtonRD/Tour.htm
• Changing the filename case works: http://sml2.som.hw.ac.uk/gb6/BroughtonRD/toUR.htm
• Changing the directory case fails!: http://sml2.som.hw.ac.uk/gb6/broughtonrd/Tour.htm

Scenario:

• One would like to be able to mount NetWare volumes or folders under Suse Linux 10.1.
• The scenario is of a single user on a Suse 10.1 machine who wishes to mount NSS volumes on a Novell NetWare 6.0 file server on which they have a regular login (eDir). The Suse environment is KDE 3.5.1 and the Novel server communications are IP only (no IPX).
• The Novell server in question is referred to as SOMFS2 but has FQDN somfs2.som.hw.ac.uk and the Suse machine is suse01 (.hw.ac.uk). The Novell user account mentioned (somdjp) is in eDir/NDS context .som.staff.hw
• All testing was performed using VMware Workstation 6.0

References:

1. A good page with discussion and lot of links to this subject is here.
   • Options: Use ncpfs
   • Novell have their own Client for Suse Linux 10 here .
2. Good website entitled /etc/fstab demystified

The Novell Client seems excellent for regular Linux desktop users, but a lighter solution is sought here.

NCPFS

The path chosen first was to install ncpfs.

1. Using Yast, install the ncpfs package (NetWare Core Protocol File System) - download site http://freshmeat.net/projects/ncpfs/
2. Novell Cool Solutions: "HOWTO: Mount Netware Servers in Linux Through /etc/fstab" [Ref]
   • Using Yast, package ncpfs was installed.
   • > mkdir /mnt/somfs2 [as root, to create the mount point]
   • > groupadd nwaccess [group to whom NetWare access is granted]
   • > usermod -G nwaccess somdjp [add the user somdjp to the nwaccess group]
   • > cd /mnt/somfs2
   • > chgrp nwaccess . [Important - see here for why.]
3. Edit /etc/fstab to include the last line shown here in blue (followed by a blank line):
   • /dev/sda2 / reiserfs acl,user_xattr 1 1
     /dev/sda1 swap swap defaults 0 0
     proc /proc proc defaults 0 0
     sysfs /sys sysfs noauto 0 0
     debugfs /sys/kernel/debug debugfs noauto 0 0
     usbfs /proc/bus/usb usbfs noauto 0 0
     devpts /dev/pts devpts mode=0620,gid=5 0 0
     /dev/fd0 /media/floppy auto noauto,user,sync 0 0
     somfs2.som.hw.ac.uk/somdjp.som.staff.hw /mnt/somfs2 ncp uid=somdjp,gid=nwaccess,mode=660,owner=somdjp,A=somfs2.som.hw.ac.uk,passwdfile=/etc/ncp-pass 0 0
4. Mount the NetWare volumes via
   • > su [must change to root to use mount command]
   • > mount -a -t ncp [password is read from /etc/ncp-mount]
   • > ls /mnt/somfs2 [all Novell SOMFS2 volumes appear under here!]
5. The /etc/ncp-pass file must be manually created and have the format
   • somfs2.som.hw.ac.uk/somdjp.som.staff.hw:mypassword
     namely FQDN of host/FQ Login name of user:password
   • Permissions on /etc/ncp-pass are -rw——- root root (so only root can access this file).
   • Update: This was later changed so that user somdjp was the owner, and the group was set as nwaccess.
6. Dismount NetWare volumes via
   • > umount -a -t ncpfs [note the last word is not ncp but ncpfs.]
7. [NB: This next part didn't work for me.]
   If non-root users are to use the mount command, one must SetUseridRoot the Mount command (see here on a related IPX mount-related web page which has some good information on mounting NetWare volumes). One would change the permissions on the mount command via
   • > whereis mount [answer is /bin/mount]
   • > chown root:nwaccess /bin/mount
   • > chmod 755 mount [owner (root) has twx, group (nwaccess) has rx, other has rx]

   Alas, this didn’t work. Issuing the correct mount -a -t ncp command as a non-root user produces no result. Perhaps the /etc/ncp-pass file permission/ownership are at issue?
   Change owner to the non-root user who wishes to mount the volume:

   • >chown somdjp /etc/ncp-pass
   • >chmod 750 /etc/ncp-pass [owner=r,w,x,group=r,x]
8. The above didn’t work either. The solution is to use the sudo command.
9. Sudo allows one to run selected commands with root (or other user) privileges. It is configured by means of the visudo command. Usage: As root, configure sudo by running
   • > visudo
10. This starts the microemacs text editor to edit the /etc/sudoers file, which it creates if not present. Note that one should set the EDITOR environment variable or vi will start. Do this by editing /root/.bashrc and adding the line
    export EDITOR=ue
11. So, back in the visudo environment, enter the following text:
    • %users suse01 = NOPASSWD: binmount -a -t ncp, sbinmount -a -t ncp, binumount -a -t ncpfs
12. This will allow all users of machine suse01 to use the mount/umount commands with the specified parameters, required to handle mounting of Novell volumes. For the user (somdjp) to mount the volumes he must run
    • > sudo mount -a -t ncp
13. To dismount volumes, one can run (as root) either of these commands:
    • > umount -a -t ncpfs
    • > umount /mnt/somfs2 [where the latter is the mount point]
      Note that using only the first command sometimes resulted in a ‘file system is busy’ warning message but using the second one always worked. Also, one could run the sync command to flush the file system buffers before dismounting.
14. It would be convenient to provide a desktop shortcut to a script which would do the mounting and dismountig, and this is covered next.

Scripting: Mounting the Novell Folders with help from kdialog

So, we can mount the NetWare folders as root, from a command shell by the above mount command.

We could script this and create a simpler bash shell script to do this to avoid remembering the mount syntax. Wouldn’t it be nice, also, to have a KDE desktop shortcut from which we could run this BASH script, which need only prompt for the ROOT password to run the mount command? The solution is using kdialog which will provide a dialogue to the user (as the name suggests.)

References: Shell scripting with KDE Dialog , Kdialog Dialog Types ,

Usage: Refer to the above for further details. The script requires to request the user for the ROOT password, so one would use the examples in the links above to:

1. In the user’s home folder, create a bash script with microemacs (for example) named, for simplicity, novell_mount.sh
   • > ue novell_mount.sh&
2. The bash script is as follows:
   #!bin/bash
   # Define the home folder of user on Novell server.
   VOL=/mnt/somfs2/somfs2/users2/somdjp

   # Tell the user what’s happening:
   kdialog –title "Novell Folder Mounting" –passivepopup "Mounting SOMFS2 Volumes …" 1

   if [ -d $VOL ]
   then
   kdialog –title "Novell Folder Mounting" –msgbox "It’s ok!nNovell Volumes are already available, dude, under $VOL ."
   exit
   else
   # Mount the volume using sudo (edit with visudo command)
   sudo mount -a -t ncp
   fi

   # Now test if home folder has mounted
   if [ -d $VOL ]
   then
   kdialog –title "Novell Folder Mounting" –passivepopup "Novell Volumes Mounted under $VOL." 2
   else
   kdialog –title "Novell Folder Mounting" –msgbox "Novell Volumes did NOT mount.n Oh dear. Try again."

   fi
   exit

3. To make the script executable one must of course type
   • > chmod u+x novell_mount.sh
4. And to run the script from a command shell one would type
   • > . novell_mount.sh
5. One can easily create a KDE desktop shortcut to this script, as long as the command parameter is
   • /bin/bash -i /home/<user>/novell_mount.sh
   • One needs also click Properties-Application-Advanced and select the "Run in Terminal" option or a bouncing ball appears for a good few seconds after the script/shortcut has run. This indicates the application times out, or does not close properly.
   • The "Terminal Options" are the KDE Konsole Options, listed here.
   • The Terminal options chosen were –vt_sz 2×1 –nomenubar –noscrollbar -T Info
6. To create a dismount script, open microemacs (or any editor) and create the following under script novell_umount.sh:
   #!bin/bash
   # Define the home folder of user on the Novell server:
   VOL=/mnt/somfs2/somfs2/users2/somdjp

   # Tell the user what’s happening:
   kdialog –title "Novell Folder Mounting" –passivepopup "Dismounting SOMFS2 Volumes …" 1

   if [ ! -d $VOL ]
   then
   kdialog –title "Novell Folder Dismounting" –msgbox "It’s ok!nNovell Volumes are not mounted."
   exit
   else
   # Flush any caches to disk
   sudo sync

   # Dismount by mount name
   sudo /bin/umount /mnt/somfs2

   # Dismount the volume type using sudo (edit with visudo command)
   sudo /bin/umount -a -t ncpfs
   fi

   # Now test if home folder has dismounted
   if [ ! -d $VOL ]
   then
   kdialog –title "Novell Folder Dismounting" –passivepopup "Novell Volumes Dismounted." 2
   else
   kdialog –title "Novell Folder Dismounting" –msgbox "Novell Volumes did NOT dismount.n Oh dear. Try again."

   fi

   exit

7. Again, create a KDE shortcut to this script and ensure it is executable, as above. Note that
   • A few more sudo commands have been added, for which visudo must be run, and the following line must be present in /etc/sudoers as a result:
   • %users suse01 = NOPASSWD: /bin/sync, /bin/mount -a -t ncp, /sbin/mount -a -t ncp, /bin/umount /mnt/somfs2, /bin/umount -a -t ncpfs
   • In the discount script, both ways of dismounting were required to ensure the volumes are dismounted. At this stage, I’m unsure why this should be. Volume dismount isn’t actually critical for single-user Linux machines.

Initial Error when running Kdialog

An error message was generated initially when using kdialog, stating "failed to open file /usr/lib/qt3/etc/settings/qtrc". A web-search revealed this was because the above file (qtrc) was readable only by root. The error was resolved by chmod a+x the above file (as root).

Mounting at Boot time: Issues

The Novell volumes are not mounted at boot time, as the network service has not started at the time /etc/fstab is being read and disks/volumes mounted. This isn’t a problem in the scenario here and is further discussed in the NetWare Cool Solutions "Automatically Mounting Remote NCP (NetWare) Shares on SUSE Linux System Boot" [Ref]
which goes into more detail and discusses Samba solutions too. Samba
would mount the volume in a case-insensitive way, suitable for the
documentroot of an Apache webserver.

The Novell article above states:

The Problem:

Many users want to have remote NCP shares mounted on their SUSE Linux machine on system boot. The ncpmount (8) man page explains that to mount a NetWare share on system boot one simply needs to add the share to /etc/fstab. They do this, but the NCP shares are not mounted on system boot.

The Cause

The cause of this problem is simple: the system is attempting to
mount the NCP shares prior to starting the networking. This obviously
fails. After the system has boot, the user can execute mount -at ncpfs as root and have the shares mounted fine

The Solution

The most simple solution is to check whether the server hosting the
NCP share supports Native File Access for SMB/CIFS. If so, the share
can be mounted via Samba. Samba has scripts which mount the shares after networking has been started. These shares can be added to either /etc/fstab or /etc/samba/smbfstab.

So, the best way is to mount the Novell volumes as Samba shares - which would also be case-insensitive and thus suitable for being mounted as the Apache htdocs folder.

Addendum: An article on Linux Clustering advises:

"Next, edit /etc/fstab to make the server mount automatically when rebooting. When doing this, make sure that the file system has the option _netdev added in the options column. This option ensures that the file system is mounted only after the network has been enabled."

Adding the parameter _netdev to /etc/fstab and rebooting caused Suse to boot into command-line mode only, requesting a username and password. Loading microemacs from the command-prompt (ue) and commenting out the whole line by prepending a # symbol at the start resolved the problem.

Later, this was changed back to the line being active (comment removed) but the password file option is removed and replaced by option noauto. This allows the user (in this case somdjp) to manually mount the Novell folders from a command shell, if desired, but they won’t be auto-mounted at system boot time. However, the NetWare folders were no longer mounted for some unknown reason (using >mount -a -t tcp). Solution: put it back to the way it was before it got broken :o)

Using ncpmount

It’s also possible to mount Novell volumes using the ncpmount command that comes as part of the ncpfs package. Initial steps in testing this are as follows:

1. In a command shell, change to root (su)
2. Make sure the mount point exists (e.g. /mnt/somfs2)
3. Issue the command
   • >ncpmount -S somfs2 -A somfs2.som.hw.ac.uk -U somdjp.som.staff.hw /mnt/somfs2

   whereupon a password will be prompted for and the Novell volumes mount.

4. There are many options available (see the hyperlink to ncpmount above). Some of these can store the password in a file, for instance.
5. To dismount the volumes one can issue
   • >ncpumount -S somfs2

Note that the normal Linux user (e.g. somdjp) does not have permissions to view the volumes (only root does) when the above command is issued. The remedy is to use the -u and -g flags which specify the default file owner (via uid) and group id of the mounted volumes. These can be determined easily by running

• > cat /etc/passwd [obtain the uid and gid of the user for whom access is required]
• > ncpmount -S somfs2 -A somfs2.som.hw.ac.uk -U somdjp.som.staff.hw -u 1002 -g 100 /mnt/somfs2

This time the regular Linux user with uid 1002 can browse their Novell files - and long filenames are supported.

Samba

Documentation

• See this Novell Course which offers a tutorial to accessing Novell volumes via Samba.
• Suse Samba Howto - video tutorial.
• How To Samba With Suse 10.1 And Windows XP - step by step guide.

Overview: PKK Pair

On the suse-somac8 Virtual Machine (Suse 10.1, KDE 3.5.4, 6Gb) a pair of public-private keys was created for user somdjp to allow secure remote connection and transfer of files.

Reference: Suse Linux 10 Bible p428, great ssh link.

Steps:

1. Login as somdjp and type
   • > ssh-keygen -t dsa (and use a passphrase - hint: same as Duncan always uses)
2. This creates folder ~/.ssh where ~ represent the home folder of somdjp, under which are found the public key (id_dsa.pub) and private key (id_dsa).
3. On any machine to which somdjp wishes to login securely the folder ~/.ssh/authorized_keys must contain the public key; both keys were copied to a USB memory stick.
4. On host suse-somac8, for which the root user is not going to be somdjp but somac8, the private key was deleted from the somdjp/.ssh home folder for security. The private key must always be - yes - private.
5. Powering up host suse01, the main VM for this Suse project, account somdjp had the PPK pair copied across.
   NB: VMware Workstation 6.0 cannot use bridged networking for two guests at the same time, the network of one will not be operational.