Sharing, Backup and Synchronisation under Ubuntu Linux

Home

This page has been extended from my original page covering standard backing up and now covers the prerequisites of Mounting Drives, File Sharing between Computers running both Linix and Windows, secure access to remote machines and Synchronisation. It also covers another technique in the armory for preservation of data, namely separation of data and system.

Introduction
Preliminaries - Background on the Linux File System
Permanently Mounting a shared drive in Ubuntu - use of fstab (Advanced)
Accessing Files on Windows Machines over the Windows Network
Accessing Files on the Linux machines from Windows- samba
Backing up

Simple Backup - sbackup utility
Use of Tar Archives to backup - tar and the archive manager

Synchronisation between machines and hard drives - Unison (Advanced)

Secure Synchronisation over a Network - using ssh (Advanced)

Moving an Ubuntu Home Folder to a dedicated Partition (Expert level)

Introduction

The ability to Backup the system has always been considered an essential to any system. The amount of data involved has increased dramatically from the days when data was preserved on floppy disks and our internal drives were 850 Mbytes (which was adequate for a Windows 95 system with Office Office 95 - now a minimum size Windows XP or Ubuntu Linux system is about 6 Gbytes and even a 5 year old laptop has a 40 Gbyte drive . Our photographs occupy 40 Gbytes and Audio 15 Gbytes - videos are not practical to back up and occupy some 750 Gbytes.

Backing up takes many forms now that the norm is for machine to be networked so this article covers many forms of achieving redundancy and preserving data. It covers shared drives between different operating systems on dual booted computer, networking of both Linux and Windows machines, techniques to separate system and data areas, conventional backing up to internal and external drives and most important in this time of mobility - synchronisation both between machines and to external hard drives.

Preliminaries - Background on the Linux File System

I am not going to extol the virtues of Linux, in particular the Ubuntu Distribution here, but I need to explain a little about the ways that file systems and disks differ in Windows and Linux before talking about backing up. In Windows, Physical Disk drives and the 'virtual' Partitions they are divided up into show up in the top level of the File System as Drives with names such as C: , a floppy disk is almost always A: and the system drive is C: . This sort of division does not appear at all in Linux where you have a single file system starting from root ( / ) which is where the operating system is located booted. Any additional disk drives which are present or are 'mounted' at a latter time such as a USB disk drive will by default in most distributions appear in the file tree in /media and the naming will depend on several things but expect to entries such as /media/cdrom , /media/floppy and /media/disk. If you create a special partition for a different part of the filesystem such as /home where all the users live then it can be 'mounted' at /home. In theory you could mount partitions for all sorts of other parts of the file system. If, for example, you add a new disk and choose to mount a partition to just contain your home directory it is only the addition of a single line in a file although you need to exactly copy the old contents to the new partition first - I will cover that in detail latter in this article.

There is a nearly perfect separation of the kernel, installed programs and users in Linux. The users each have a folder within the home folder which has all the configuration of the programs they use which is specific to them - it is set up the first time they run a program. A users folder only contains a few tens of kbytes until programs are used and all the program settings are hidden (hidden files and folders start with a . (dot). There are usually a number of folders generated by default including /Desktop /Documents /Music /Pictures /Videos /Templates and /Public which are used by program as their defaults. This makes backing up very easy - an exact copy of the home folder allows the system to be restored after a complete reload of the operating system and programs. Note the word exact as the 'copy' has to preserve Symbolic Links and the Permissions of all the files - permissions are key to the security of Linux so special archiving utilities are best employed.

Permanently Mounting a shared drive in Ubuntu 8.04 and 9.04 (Advanced)

If we are going to use a shared drive for data then we must ensure that it is permanently mounted. The mounting points in all major flavours of Linux are defined in a file system table which is in /etc/fstab and and the convention is that the mount points are in /media. We therefore need to set modify /etc/fstab to set up to mount points in /media and we must also create the directories for them using sudo to make them owned by root and set the permissions so they are accessible to all users.

I take a simple approach and open the File Browser as Root in a terminal by

gksudo nautilus

This allows me to use the graphical file browser to create the folders and set permissions by standard navigation and right click menus for create folder and properties -> permisions. It is best to make these folders with the same names as those assigned by mounting from 'Places' which is derived from the partition label if it is set. Do not continue to use this Root File Browser after setting up the shared folder as running anything as Root has many dangers of accidental damage although the aware reader will realise that the terminal can be avoided in the next stage by also opening fstab from within the Root File Browser - but do take care!

It is desirable to back up /etc/fstab and then make the changes using the editor gedit. This is done in a terminal by:

sudo cp /etc/fstab /etc/fstab_backup
sudo gedit /etc/fstab

The procedure for modifying /etc/fstab in Hardy Heron and Jaunty is the same as it was for Dapper except that it is now recommended to use the drives UUID rather than the device's location ie append lines to /etc/fstab looking like:

# /dev/sda5
UUID=49D3-6030 /media/DATA vfat iocharset=utf8,umask=000 0 0
# /dev/sda2
UUID=F6EC309DEC305A57 /media/WINDOWS ntfs nls=utf8,umask=0222 0 0

Note – The first string is for a FAT32 partitions and the second for a NTFS partitions. The first provide a read/write mounts ( umask=000 for read/write for owner, group and everyone else) and the second a read only mount (umask=0222). If you want to only allow others read access then use umask=002

The UUID can be found by typing sudo blkid in a terminal – typical output looks like:

pcurtis@vortex-ubuntu:~$ blkid
/dev/loop0: TYPE="squashfs"
/dev/sda1: LABEL="WINRE" UUID="80AE-9D55" TYPE="vfat"
/dev/sda2: UUID="F6EC309DEC305A57" LABEL="WINDOWS" TYPE="ntfs"
/dev/sda5: LABEL="DATA" UUID="49D3-6030" TYPE="vfat"
/dev/sda6: UUID="8993fcc3-6a5a-4f94-9e6a-c77f5091c1ad" TYPE="swap"
/dev/sda7: UUID="83ac79eb-c78f-42a8-bbde-60ef0145a289" TYPE="ext3"
/dev/sda8: UUID="e31b581a-85fa-4681-bf11-aebb9a9656cd" TYPE="ext3"
pcurtis@vortex-ubuntu:~$

After modifying /etc/fstab and rebooting the three Windows Drives are mounted and appear on the desktop in addition to being in 'Places' - they can not be unmounted without root privileges which is just what we want.

There is one 'feature' of this way of mounting which seems to be totally universal and that is that only root (the owner) can set the time codes - this means that any files or directories that are copied by a user have the time of the copy as their date stamp. What seems to happen is this:

When copying a file to a fat partition, it gets - during its creation - the default (or defined) 'virtual' user, group ids, attributes and the current time stamp.
Only the owner of a file or root can change its time stamp.
This means, if your own user id does not match the 'virtual' user-id, you can not change the file date back to file date the source file after a copy.

A solution for a single user machine is to find out your user id and mount the partition with option uid=user-id, then all the files on that partition belong to you - even the newly created ones. This way when you copy you keep the original file date. This is important if you have a file synchronisation program such as Unison which checks file creation and modification dates.

# /dev/sda5
UUID=706B-4EE3 /media/DATA vfat iocharset=utf8,uid=yourusername,umask=000 0 0

You must change yourusername to you own user name.

In the case of multiple user machines you should not mount at boot time and instead mount the drives from Places.

The uid can also be specified numerically and the first user created has user id 1000.

Accessing Files on Windows Machines over the Windows Network

I originally considered this would be an important step but the order of difficulty was not what I expected - I could access files on Windows machines over the Network immediately using Places -> Network Servers although the creation date is only displayed correctly for files on NTFS partitions. I have not found a workaround yet or any reference to it despite extensive web searches. You will usually be asked for a password the first time.

Accessing Files on your Linux machine from Windows machines (and other Linux machines on your local network).

Install Samba: A file server to be installed on the Linux machine called Samba (Using Add/Remove as usual) System -> Administration -> Samba this gives access to a control panel for setting up the shares, workgroup, users, passwords etc.
Set Workgroup: the workgroup used by all the machines on the Network is set by System -> Administration -> Samba -> Preferences -> Server Settings -> Basic tab and enter the workgroup name used by the Windows Network.
Provide a Username and Password for login from other machines: You the need to set up a username and password to use from the Windows and other machines when you log into this machines in the Samba Control Panel by: System -> Administration -> Samba -> Preferences -> Samba Users and add/edit a user to provide a username and password (I use the same as on my main Ubuntu user but you may wish to use a different one if others use the Windows machines) for Windows Networking Users to log into the machine. NB the password must be set even if you use an existing user name.
Set up Shared Folders: You can now set up the shares you want by Add Share - the screens are clear in their requirements.

All these settings need a restart to become active. To save a full restart you can just restart Samba by typing on a terminal:

sudo /etc/init.d/samba restart

You can now use e Network Browser on another machine to access the machine you have just set up by Places -> Network - you will be asked for the Username and Password you have just set up and you can specify how long the password remains active.

For information: There are other ways of file sharing between Linux machines but samba is the popular choice as it also covers Windows machines.

Backing up.

Simple Backup (sbackup).

There are a number of backup utilities available for Ubuntu Linux but use Simple Backup which is a simple backup solution intended for desktop use created within Google Summer of Code 2005 for Ubuntu with the mentoring of the Ubuntu team. It is a tiny download and seems very comprehensive in the ways you can chose to include and exclude files based on location, size, type etc. It can run automatically and will do periodic full backups with incremental backups of changed files. By default it saves the backup as a compressed file in /var/backup and the file seems very small at under 400 Mbytes with default settings. You can restore individual files etc and in theory they can be restored to a different location although that feature has not worked for me. You can also access the various levels of archive folders and use archive manager to extract directly from the .tar archives in the folders.

It is all very easy with a good GUI interface and is found in Systems ->Administration -> Simple Backup. Simple Backup is downloaded by Add/Remove and search for Simple Backup or Synaptic package manager and search for sbackup.

The default settings create a daily incremental backup and a weekly full backup. Although the files are quite small you need to have a policy of a monthly tidy up otherwise you will end up with a full disk - it has happened to me twice

Standard Tar Archives

There is also a very powerful command line archiving tool round which many of the GUI tools are based which should work on most Linux Distributions. In many circumstances it is best to access this directly to backup your system. The resulting files can also be accessed (or created) by the archive manager accessed by right clicking on a .tzr or .trz file. To show the power the following commands will back up your entire file system but can easily be adapted to just back up your home folder by starting there instead of in the root directory:

In a terminal first change to the root directory:

cd /

then:

sudo tar cvpzf mybackup.tgz / --exclude=/mybackup.tgz --exclude=/proc --exclude=/lost+found --exclude=/media --exclude=/sys --exclude=/mnt

This is a single line if you cut and paste. The options cvpzf are: create archive, verbose mode, retain permissions, gzip archive and file output. Then follows the name of the file to be created, mybackup.tgz which should use the date for easy reference. Next is the directory to back up, in this case / the root so everything will be stored. (Another common possibility would be /home). Next are the objects which need to be excluded - the most important of these is our back up file or it would be recursive! It also excludes directories which are recreated dynamically.

The backup process is slow (15 mins plus) and the file over a gbyte for the simplest system. After it is complete the file should be moved to a safe location, preferably a DVD or external device. If you want to do a higher compression method the command "tar cvpjf mybackup.tar.bz2" can be used in place of "tar cvpzf mybackup.tgz". This will use bzip2 to do the compressing - j option. This method will take longer but gives a smaller file.

You can access parts of the archive using the GUI Archive Manager by right clicking on the .tgz file - again slow on such a large archive. A full system backup is most useful if you can effectively restore your entire system or data. This can be again done by a few commands but do not try this for fun as it will overwrite the entire Ubuntu file system, thus restoring the older image that we took. I have not yet had problems so large I have had to try this! The following assumes the backup image is still in the root directory or has been replaced there from DVD:

cd /
sudo tar xvpfz mybackup.tgz -C /

The restoration uses the options - extract, verbose, retain permissions, from file and gzip. This will take a while because all your files will be overwritten with the versions from the image you previously backed up. The "-C /" ensures that the file is restore to the root directory /.

If this has been done onto an empty drive from a LiveCD one needs to replace any excluded directories before rebooting the system.

mkdir proc
mkdir lost+found
mkdir sys

If the old system is still present note that it only overwrites files, it does not deleted files from the old version which are no longer needed.

This is one of the simplest and fastest methods for backing up. It may take some customizing and tweaking for your purposes but it is powerful, versatile, and free.

Synchronisation between machines and hard drives - Unison (Advanced)

I have put this as an advanced activity as a number of features of the normal set up of the way removable and fixed drives are mounted need to be understood and configured to avoid unexpected results. See below under Permissions and Automounted Drives (Removable Drives) and Drives Mounted Permanently at startup using fstab.

Backing up is very important and often the easiest way has been to make copies onto external hard drives and onto other machines. But what does one do when one has multiple copies of important folders and different files have been added or edited on each machine or external hard drive copy. This is a particular problem when one uses a laptop or netbook away from home for a few months and returns especially when only subsets of files may have been taken with one or the two machines were not in sync when one started.

Linux has a very powerful tool available to synchronise folders, and all their subfolders, between drives on the same machine and across a local network using a secure transport called SSH (Safe 'S Hell). At its simplest you can use a Graphical User Interface (GUI) to synchronise two folders which can be on any of your local drives, a USB external hard drive or on a networked machine which also has Unison and SSH installed. Versions are even available for Windows machines but one should make sure that the Unison versions numbers are compatible.

You just enter or browse for the two folders if local and it will give you a list of differences and recommended actions which you can review and it is a single keystroke to change any you do not agree with. Unison uses a very efficient mechanism to transfer/update files which minimises the data flows based on a utility called rsync. The initial Synchronisation can be slow but after it has made its lists it is quite quick even over a slow network between machines because it is running on both machines and transferring minimum data - it is actually slower synchronising to another hard drive on the same machine.

For more complex synchronisation with multiple folders and perhaps exclusions you set up a very simple configuration file for each Synchronisation Profile and then select and run it from the GUI as often as you like. It is easier to do than describe - a file to synchronise my four important folders My Documents, My Web Site, Web Sites, and My Pictures is only 10 lines long and contains under 200 characters yet synchronises 25,000 files! The review list is intelligent and if you have made a folder full of sub folders of pictures whilst you are away it only shows the top folder level which has to transferred which is a good job as we often come back with several thousand new pictures!

Both Unison and SSH are available in Ubuntu but need to be installed using System -> Administration -> Synaptic Package Manager and search for and load the following three packages:

unison-gtk ssh winbind

or they can be directly installed from a terminal by:

sudo apt-get install unison-gtk ssh winbind

The first is the GUI version of Unison which also installs the terminal version if you need it. The second is a meta package which install the SSH server and Client, blacklists and various library routines. The third is required to allow ssh to use host names rather than just absolute IP addresses which will also requires some extra configuration described below - you may find with some versions of Ubuntu that winbind is already installed but it will still need the configuration.

Unison is then accessible from the Applications Menu and can be tried out immediately. There is one caution - it is initially configured so that the date of any transferred file is the current date ie the file creation date is not preserved although that can easily be set in the configuration files - the same goes for preserving the user and group.

Permissions for External USB Drives and Drives Mounted via 'Places': The file systems on External USB Drives are almost always Windows type file systems such as FAT32 which do not support the definitions of owners and groups and their associated permissions. The same applies to such file systems on internal drives which are mounted through Places or permanently mounted when the machine is started up. The Owner, Group and Permissions are therefore set up each time when the drives are mounted. Unison checks for matches in the permissions so one has to make sure that USB drives are mounted with suitable and matching permissions - by default they are mounted with 0700 whilst drives permanently mounted in the File System Table at startup usually have the permissions set to 777 or 775 in /etc/fstab. See above for information on permanently mounting drives and /etc/fstab

The permissions for internal drives mounted via Places or automounted when external drives are plugged in can and should be set using the Gnome Desktop Configuration Editor to match what is set up for permanently mounted drives in /etc/fstab. The Configuration Editor is norminally installed but does not show under Applications -> System -> Configuration Editor (because it is powerful and open to abuse by the unwary) . You can set up the system to display by a Right click on the Applications drop down -> Edit Menus and find and tick the box for it to be displayed. It is a useful program so it is worth making it easily available but if you are in a hurry it can also be run from a terminal by:

gconf-editor

Once you have the configuration editor running expand the parameters on the right through system -> storage -> default_options and set umask=000 for the drive type which will probably be vfat. Double click down till you get to the existing umask=077 and change it to 000 or 002 to match drives mounted by fstab. Note umask is a mask and you set the inverse of the permissions you want which are 777 or 775 !

Even so Unison may seems to get some miss-matchs on the first synchronisation which have to be Resolved. I have set up a number of profiles for back-up and synchronisation with my WD Passport portable 120 Gbyte USB drive which is formated to FAT32 (vfat) to keep Pauline's Toshiba Laptop and MSI Wind netbook synchronised and a Profile to Synchronise my MSI Wind Netbook and homebuilt Desktop over the local network.

Ownership of Drives Mounted permanently using fstab: There is one 'feature' of mounting using the File System Table in the usual way is that the owner is root and only the owner (root) can set the time codes - this means that any files or directories that are copied by a user have the time of the copy as their date stamp which can cause problems with Unison when Synchronising. What seems to happen is this:

When copying a file to a fat partition, it gets - during its creation - the default (or defined) 'virtual' user, group ids, attributes and the current time stamp.
Only the owner of a file or root can change its time stamp.
This means, if your own user id does not match the 'virtual' user-id, you can not change the file date back to file date the source file after a copy.

# /dev/sda5
UUID=706B-4EE3 /media/DATA vfat iocharset=utf8,uid=yourusername,umask=000 0 0

In the case of multiple user machines you should not mount at boot time and instead mount the drives from Places.

The uid can also be specified numerically and the first user created has user id 1000.

SSH (Secure aS Hell)

Using SSH to synchronise between machines over a network

So far we have looked at the set needed up for Unison to Synchronise files on the same machine. We will now look at synchronising between machines using ssh. When you set up the two 'root' directories for synchronisation you get four options when you come to the second one - we have used the local option but if you want to synchronise between machines you select the SSH option. You then fill in the hostname which can be an absolute IP address or a hostname (but see below on the need to set this up). You should then specify the username you will use to log into the other machine ie the samba sharing username as set up for windows sharing, leave the port blank. You will also need to know the folder on the other machine as you can not browse for it. When you come to synchronise you will have to give the password corresponding to that username.

Setting up ssh .

I have always checked out that ssh has been correctly set up on both machines before trying to use Unison. In its simplest form ssh allows on to log using a terminal on a remote machine. Both machines must have ssh installed (and the ssh daemons running which is the default after you have installed it). For reasons we will come to (and sort out) the set up of Ubuntu only allows one to use and absolute IP address so you need to know the address on your network of the other machine. This can be found within a mass of other information by opening a terminal on the machine whose IP address you need and typing

ifconfig

which will give a lot of information on all your connections including the active network - the IP address you are looking for will follow an inet addr: . There will be two addresses, one will be for the local loopback (127.0.0.1) and the other is the external address which is what we want.

if you want to reduce the amount of information to just the lines containg inet addr then you can pipe the output through a program called grep - I leave it as a task for the reader to find out more about pipes and grep once you have tried it and can see the power of the command line!

ifconfig | grep "inet addr"

It is now time to try ssh and I assume that the machine we want to access has an IP address 192.168.1.2

ssh 192.168.1.2

You will get some warnings that it can not authenticate the connection which is not surprising as it is the first time and will ask for confirmation and you have type yes rather than y. It will then tell you it has saved the authentication information for the future and you will get a request for a password which is the start of your log in on the other machine. After providing the password you will get a few more lines of information and be back to a normal terminal prompt but note that it is now showing the address of the other machine. You can enter some simple commands such as a directory list (ls) if you want. When you have tired of this type exit to return to your own machine.

Enabling Hostname Resolution under Ubuntu: Using absolute IP addresses is not very satisfactory in the long term as they are allocated dynamically from your router or other server on most modern networks. This means that they may change so you could end up synchronising with the wronge machine. In some ways the consistency and ssecurity checks run by ssh to prevent hacking are even more of a problem when the IP address changes and I have sometimes had to clear out the file containing that information (~/.ssh/known-hosts) . It is therefore much better to use the hostname to access the machine. Unfortunately the default set up in Ubuntu does not support hostname resolution without an additional utility being loaded and a change made in a configuration file to enable what is called WINS (Windows Internet Name Service) resolution by:

Installing winbind using System -> Administration -> Synaptic Package Manager and search for winbind - it is often installed with samba so you may already have it but it is best to check.

then in a terminal:

sudo gedit /etc/nsswitch.conf

and change this line:

hosts: files dns

to:

hosts: files wins dns

Note - in some cases there may be more on the line - if that is the case put it immediately after files.

Now ssh <hostname> will log you into <hostname> and you can use hostnames in Unison

Some sample Profiles for Unison:

The profiles live in /home/username/.unison

# Profile to synchronise from desktop triton-ubuntu to netbook vortex-ubuntu
# with username pcurtis on vortex-ubuntu
# Note: if the hostname is a problem then you can also use an absolute address
# such as 192.168.1.4 on vortex-ubuntu
#
# Roots for the synchronisation
root = /media/DATA
root = ssh://pcurtis@vortex-ubuntu//media/DATA
#
# Paths to synchronise
path = My Backups
path = My Web Site
path = My Documents
path = Web Sites
#
# Some typical regexps specifying names and paths to ignore
ignore = Name temp.*
ignore = Name *~
ignore = Name .*~
ignore = Name *.tmp
#
# Some typical Options - only times is essential
#
# When fastcheck is set to true, Unison will use the modification time and length of a
# file as a ‘pseudo inode number’ when scanning replicas for updates, instead of reading
# the full contents of every file. Faster for Windows file systems.
fastcheck = true
#
# When times is set to true, file modification times (but not directory modtimes) are propagated.
times = true
#
# When owner is set to true, the owner attributes of the files are synchronized.
owner = true
#
# When group is set to true, the group attributes of the files are synchronized.
group = true
#
# The integer value of this preference is a mask indicating which permission bits should be synchronized other than set-uid.
perms = 0o1777

The above is a fairly comprehensive profile file to act as a framework and the various sections are explained in the comments.

Moving an Ubuntu Home Folder to a dedicated Partition (Expert level)

This is based on Psychocats - Create a separate home partition in Ubuntu with quite a few modifications to get it to work for me using Ubuntu Hardy Heron. As an aside the original article is one of an excellent series which are well worth a look. The proceedures here are not very well tested as I have only done it for one machine and one should make backups of everything of value but the techniques are used in many places. Both the partitioning and the "copying" and mounting have risks but the proceedures here are designed to minimise the risk and provide some escape routes if it does not work. It is however best to think ahead and provide the separation of system and data when the initial installation is carried out.

The first stage is to make a new partition using Gparted from an Ubuntu LiveCD. This will usually involve shrinking an existing partition, probably the root partition with the home folder you are moving. It is always safest to only reduce/increase the size of a partition leaving the start position the same. In the following I have used as an example a new ext3 partition /dev/sdb8 - you need to substitute what it turns out to be for you from Gparted. I have covered Partitioning elsewhere so I will not do more here than suggest you make sure you have a sensible division leaving space for both the main system and the /home directory and enough working space so you can keep the original /home for a while until everything is debugged.

Assuming you have made a new partition you now need to go back into your normal system. If you are using Hardy heron you can mount the new partition using Places -> Removable Disk (check the mount point is /media/disk or modify the following suitably) then use cpio which is an archiving program to 'copy' your home directory to it by the following commands in a terminal:

cd /home
sudo find . -depth -print0 | sudo cpio -p0vud /media/disk

I initially thought doing this from a LiveCD would be more sensible as it would avoid problems with files being in use or changing but it does not work as the permissions end up being incorrect as the original root folder one mounts and uses for the "copy" does not have the original permissions when using the live CD. I found that out the hard way as it was not clear in the original article and I had to start again several times! It takes a while to realise that Linux is very different from Windoze and one can work on a live system without problems of files being locked and/or in use. The complexity and use of cpio instead of cp is to maintain permissions and the various links and comes from a variety of sources. It is worth doing a man cpio and a man find to understand the programs and parameters but I will try to explain some of basics of it step by step:

cd /home - this changes to the home directory.

sudo find . -depth -print0 - this provides a list of filenames. I am not sure if the sudo is desirable or essential, it is not needed for a single user system. The - printo parameter specifies "null terminated strings" which will work even if some of the files have spaces, newlines, or other dubious characters in them. The -depth parameter specifies processing each directory’s contents before the directory itself.

The results are written into a pipe by the | and the program reading them must be capable of using this list. The cpio archiving command has this feature.

| cpio -p0vd /media/disk - here's the tricky part. This uses the "passthrough" mode of cpio, an archiving program that normally copies files "in" or "out" but it can do "both" using this "passthrough" mode. The -p sets the "passthrough" mode for cpio which then expects a list of filenames on its standard input (which we are providing from the 'find' command). It then copies the corresponding file "in" from the path specified (as part of the input line) and "out" to the the path specified as the final one of cpio's arguments (/media/disk in this case).

The rest of the switches on this cpio command are: -0 - expect the input records (lines) to be null terminated, -v provides verbose output, and -d - make leading directories as needed. -u forces overwriting of any existing files but it is best to make sure that everything is deleted before starting.

We sould now have an exact "copy" of the /home folder on the new partition with the correct permissions (at least for the username you used when carrying out the cpio copy). It is prudent to mount the partition and have a look to see if everything is reasonable. It should be [still] be visible in Places -> Removable Media for you to check the permissions and ownership is correct for each user.

Now we need to mount this in the correct place at turn on. Automatic mounting is carried out by backing up then modifying /etc/fstab by:

sudo cp /etc/fstab /etc/fstab_backup
sudo gedit /etc/fstab

and adding an extra line at the end

/dev/sdb8 /home ext3 nodev,nosuid 0 2

In Hardy Heron one ought to use the UUID making the addition into two lines

# /dev/sdb8
UUID=3b50dbce-28c8-46fc-bc63-f89bb06c54e5 /home ext3 nodev,nosuid 0 2

where your UUID can be found by:

sudo blkid

At the next start-up the modified fstab will mount the new home folder from the disk partition over the top of the existing one which is then invisible. A return to the old set up is as simple as restoring the original fstab or commenting out the new final line at which point the old /home is once more visible. At a point in the future when one is completely happy the old contents can be deleted to save space using the LiveCD.

An alternative is to rename /home to /home-backup using the LiveCD and create a new /home ready for the mounting with the appropriate permissions before changing fstab as in the original article. This is more complex and risky than mounting over the top but the 'backup' /home directory can be deleted without having to use the LiveCD.

The proceedure above was what finally worked for me with a single user setup, it is possible that permissions may need to be set up for other users. If the permissions are wrong, as happens if you do the cpio 'copy' using the LiveCD you get messages when you try to log in such as:

Users $Home/.dmrc file is being ignored. This prevents the default session and language from being saved. File should be owned by user and have 644 permissions.
Users $Home directory must be owned by user and not writable by others.

If you can get past this message, perhaps by changing permissions from the liveCD the following brute force procedure (which is a refinement of what I did in one of my early attempts) to set the groups and permissions can be tried on the other users:

sudo chown -R eachusername:eachusername /home/eachusername
sudo chmod -R 755 /home/eachusername
sudo chmod 644 home/eachusername/.dmrc

This needs to be done before each username is logged into the first time otherwise many desktop settings will be lost at the point the warning message appears - this means that the cpio 'copy' will need to be repeated and the permissions set for each additional user as above before logging in the first time.

The best way I believe now is to create a new user just before making the change and logging into that user to carry it out. The overhead in size of a the new user's directory in /home is only about 28K and it makes sure that there is very little going on and no important files are missed or changed during the copy - I worry about programs with Daemons running such as Picasa and Skype. The new user can be removed after the change has taken place. The new user is easy to creat by System -> Administration -> Users and Groups. The new user needs a name, to have the Profile of an Administrator set in the box and the passwords input, nothing else matters and can be empty or default. The new user will automatically have the ability to Sudo if set up with the profile of an Administrator.

When you delete the user using the GUI you do not delete the /home directory or another few files - to also remove the home directory it is easiest to instead use the command line and the deluser utility:

sudo deluser --remove-home username

Alternatively all files owned by the user, including those in the user's home directory may be removed as follows:

sudo deluser --remove-all-files username

One possible improvement which I have not tried would be to use rsync instead of cpio to do the intelligent copy by

rsync -avx /home/ /media/disk

you do not need to be in the home directory. rsync is an advanced archiving and copying tool which can work over ssh (Secure Shell) to other machines as just a simple intellegent copy. It is also useful for regular 'backups' as it will only copy files which have changed so it can be run on a regular basis to syncronise two machines although unison (which uses rsync) is a much better way to go.

In summary: there are many advantages in having ones home directory on a separate partition but overall this change is not a proceedure to be carried out unless you are prepared to experiment a little. It is much better to get it right and create one when installing the system.