|
|
Table of Contents
Stackable VFS (Virtual File System) modules support was new to Samba-3 and has proven quite popular. Samba passes each request to access the UNIX file system through the loaded VFS modules. This chapter covers the modules that come with the Samba source and provides references to some external modules.
If not supplied with your platform distribution binary Samba package, you may have problems compiling these modules, as shared libraries are compiled and linked in different ways on different systems. They currently have been tested against GNU/Linux and IRIX.
To use the VFS modules, create a share similar to the one below. The important parameter is the vfs objects parameter where you can list one or more VFS modules by name. For example, to log all access to files and put deleted files in a recycle bin, see the smb.conf with VFS modules example:
Example 22.1. smb.conf with VFS modules
[audit] |
comment = Audited /data directory |
path = /data |
vfs objects = audit recycle |
writeable = yes |
browseable = yes |
The modules are used in the order in which they are specified. Let's say that you want to both have a virus scanner module and a recycle bin module. It is wise to put the virus scanner module as the first one so that it is the first to get run and may detect a virus immediately, before any action is performed on that file. vfs objects = vscan-clamav recycle
Samba will attempt to load modules from the /lib directory in the root directory of the
Samba installation (usually /usr/lib/samba/vfs or
/usr/local/samba/lib/vfs).
Some modules can be used twice for the same share. This can be done using a configuration similar to the one shown in the smb.conf with multiple VFS modules.
Example 22.2. smb.conf with multiple VFS modules
[test] |
comment = VFS TEST |
path = /data |
writeable = yes |
browseable = yes |
vfs objects = example:example1 example example:test |
example1: parameter = 1 |
example: parameter = 5 |
test: parameter = 7 |
A simple module to audit file access to the syslog facility. The following operations are logged:
share
connect/disconnect
directory opens/create/remove
file open/close/rename/unlink/chmod
This module is identical with the audit module above except
that it sends audit logs to both syslog as well as the smbd log files. The
log level for this module is set in the smb.conf file.
Valid settings and the information that will be recorded are shown in the next table.
Table 22.1. Extended Auditing Log Information
| Log Level | Log Details - File and Directory Operations |
|---|---|
| 0 | Make Directory, Remove Directory, Unlink |
| 1 | Open Directory, Rename File, Change Permissions/ACLs |
| 2 | Open & Close File |
| 10 | Maximum Debug Level |
This auditing tool is more felxible than most people readily will recognize. There are a number of ways by which useful logging information can be recorded.
Syslog can be used to record all transaction. This can be disabled by setting
in the smb.conf file syslog = 0.
Logging can take place to the default log file (log.smbd)
for all loaded VFS modules just by setting in the smb.conf file
log level = 0 vfs:x, where x is the log level.
This will disable general logging while activating all logging of VFS
module activity at the log level specified.
Detailed logging can be obtained per user, per client machine, etc.
This requires the above together with the creative use of the
log file settings.
An example of detailed per-user and per-machine logging can be obtained by setting log file = /var/log/samba/%U.%m.log.
Auditing information often must be preserved for a long time. So that the log files do not get rotated
it is essential that the max log size = 0 be set
in the smb.conf file.
This module was created to allow Roaming Profile files and directories to be set (on the Samba server under UNIX) as read only. This module will, if installed on the Profiles share, report to the client that the Profile files and directories are writeable. This satisfies the client even though the files will never be overwritten as the client logs out or shuts down.
A Recycle Bin-like module. Where used, unlink calls will be intercepted and files moved to the recycle directory instead of being deleted. This gives the same effect as the Recycle Bin on Windows computers.
The Recycle Bin will not appear in Windows Explorer views of the
network file system (share) nor on any mapped drive. Instead, a directory called .recycle
will be automatically created when the first file is deleted. Users can recover files from the
.recycle directory. If the recycle:keeptree has been specified,
deleted files will be found in a path identical with that from which the file was deleted.
Supported options for the recycle module are as follow:
Relative path of the directory where deleted files should be moved.
Specifies whether the directory structure should be kept or if the files in the directory that is being deleted should be kept separately in the recycle bin.
If this option is set, two files
with the same name that are deleted will both
be kept in the recycle bin. Newer deleted versions
of a file will be called “Copy #x of filename”.
Specifies whether a file's access date should be touched when the file is moved to the recycle bin.
Files that are larger than the number of bytes specified by this parameter will not be put into the recycle bin.
List of files that should not be put into the recycle bin when deleted, but deleted in the regular way.
Contains a list of directories. When files from these directories are deleted, they are not put into the recycle bin but are deleted in the regular way.
Specifies a list of paths (wildcards such as * and ? are supported) for which no versioning should be used. Only useful when recycle:versions is enabled.
A netatalk module will ease co-existence of Samba and netatalk file sharing services.
Advantages compared to the old netatalk module:
Does not care about creating .AppleDouble forks, just keeps them in sync.
If a share in smb.conf does not contain .AppleDouble item in hide or veto list, it will be added automatically.
THIS IS NOT A BACKUP, ARCHIVAL, OR VERSION CONTROL SOLUTION!
With Samba or Windows servers, shadow_copy is designed to be an end-user tool only. It does not replace or enhance your backup and archival solutions and should in no way be considered as such. Additionally, if you need version control, implement a version control system. You have been warned.
The shadow_copy module allows you to setup functionality that is similar to MS shadow copy services. When setup properly, this module allows Microsoft shadow copy clients to browse "shadow copies" on Samba shares. You will need to install the shadow copy client. You can get the MS shadow copy client here.. Note the additional requirements for pre-Windows XP clients. I did not test this functionality with any pre-Windows XP clients. You should be able to get more information about MS Shadow Copy from the Microsoft's site.
The shadow_copy VFS module requires some underlying file system setup with some sort of Logical Volume Manager (LVM) such as LVM1, LVM2, or EVMS. Setting up LVM is beyond the scope of this document; however, we will outline the steps we took to test this functionality for example purposes only. You need to make sure the LVM implementation you choose to deploy is ready for production. Make sure you do plenty of tests.
Here are some common resources for LVM and EVMS:
See Learning Linux LVM, Part 1 and Learning Linux LWM, Part 2 for Daniel Robbins' well-written, two part tutorial on Linux and LVM using LVM source code and reiserfs.
At the time of this writing, not much testing has been done. I tested the shadow copy VFS module with a specific scenario which was not deployed in a production environment, but more as a proof of concept. The scenario involved a Samba-3 file server on Debian Sarge with an XFS file system and LVM1. I do NOT recommend you use this as a solution without doing your own due diligence with regard to all the components presented here. That said, following is an basic outline of how I got things going.
Installed Operating System . In my tests, I used Debian Sarge (i.e., testing) on an XFS file system. Setting up the OS is a bit beyond the scope of this document. It is assumed that you have a working OS capable of running Samba.
Install & Configure Samba. See the installation section of this HOWTO for more detail on this. It doesn't matter if it is a Domain Controller or Member File Server, but it is assumed that you have a working Samba 3.0.3 or later server running.
Install & Configure LVM. Before you can make shadow copies available to the client, you have to create the shadow copies. This is done by taking some sort of file system snapshot. Snapshots are a typical feature of Logical Volume Managers such as LVM, so we first need to have that setup.
The following is provided as an example and will be most helpful for Debian users. Again, this was tested using the "testing" or "Sarge" distribution.
Install lvm10 and devfsd packages if you have not done so already. On Debian systems, you are warned of the interaction of devfs and lvm1 which requires the use of devfs filenames. Running apt-get update && apt-get install lvm10 devfsd xfsprogs should do the trick for this example.
Now you need to create a volume. You will need to create a partition (or partitions) to add to your volume. Use your favorite partitioning tool (e.g., Linux fdisk, cfdisk, etc.). The partition type should be set to 0x8e for "Linux LVM." In this example, we will use /dev/hdb1.
Once you have the Linux LVM partition (type 0x8e), you can run a series of commands to create the LVM volume.
You can use several disks and/or partitions, but we will use only one in this example. You may also need to
load the kernel module with something like modprobe lvm-mod and set your system up to load
it on reboot by adding it to (/etc/modules).
Create the volume group and add /dev/hda1 to it with vgcreate shadowvol /dev/hdb1
You can use vgdisplay to review information about the volume group.
Now you can create the logical volume with something like lvcreate -L400M -nsh_test shadowvol
This creates the logical volume of 400 MBs named "sh_test" in the volume group we created called shadowvol.
If everything is working so far, you should see them in /dev/shadowvol.
Now we should be ready to format the logical volume we named sh_test with mkfs.xfs /dev/shadowvol/sh_test
You can format the logical volume with any file system you choose, but make sure to use one that allows you to take advantage of the additional features of LVM such as freezing, resizing, and growing your file systems.
Now we have an LVM volume where we can play with the shadow_copy VFS module.
Now we need to prepare the directory with something like
root# mkdir -p /data/shadow_share
or whatever you want to name your shadow copy-enabled Samba share. Make sure you set the permissions so that you can use it. If in doubt, use chmod 777 /data/shadow_share and tighten the permissions once you get things working.
Mount the LVM volume using something like mount /dev/shadowvol/sh_test /data/shadow_share
You may also want to edit your /etc/fstab so that this partition mounts during the system boot.
Install & Configure the shadow_copy VFS Module. Finally we get to the actual shadow_copy VFS module. The shadow_copy VFS module should be available in Samba 3.0.3 and higher. The smb.conf configuration is pretty standard. Here is our example of a share configured with the shadow_copy VFS module:
Create Snapshots and Make Them Available to shadow_copy.so. Before you can browse the shadow copies, you must create them and mount them. This will most likely be done with a script that runs as a cron job. With this particular solution, the shadow_copy VFS module is used to browse LVM snapshots. Those snapshots are not created by the module. They are not made available by the module either. This module allows the shadow copy-enabled client to browse the snapshots you take and make available.
Here is a simple script used to create and mount the snapshots:
#!/bin/bash
# This is a test, this is only a test
SNAPNAME=`date +%Y.%m.%d-%H.%M.%S`
xfs_freeze -f /data/shadow_share/
lvcreate -L10M -s -n $SNAPNAME /dev/shadowvol/sh_test
xfs_freeze -u /data/shadow_share/
mkdir /data/shadow_share/@GMT-$SNAPNAME
mount /dev/shadowvol/$SNAPNAME \
/data/shadow_share/@GMT-$SNAPNAME -onouuid,ro
Note that the script does not handle other things like remounting snapshots on reboot.
Test From Client. To test, you will need to install the shadow copy client which you can obtain from the Microsoft web site. I only tested this with an XP client so your results may vary with other pre-XP clients. Once installed, with your XP client you can right-click on specific files or in the empty space of the shadow_share and view the "properties." If anything has changed, then you will see it on the "Previous Versions" tab of the properties window.
This section contains a listing of various other VFS modules that have been posted but do not currently reside in the Samba CVS tree for one reason or another (e.g., it is easy for the maintainer to have his or her own CVS tree).
No statements about the stability or functionality of any module should be implied due to its presence here.
URL: Taylors University DatabaeFS
I have created a VFS module that implements a fairly complete read-only filesystem. It presents information from a database as a filesystem in a modular and generic way to allow different databases to be used. (Originally designed for organizing MP3s under directories such as “Artists,” “Song Keywords,” and so on. I have since easily applied it to a student roster database.) The directory structure is stored in the database itself and the module makes no assumptions about the database structure beyond the table it requires to run.
Any feedback would be appreciated: comments, suggestions, patches, and so on. If nothing else, it might prove useful for someone else who wishes to create a virtual filesystem.