SMBTA

Revision as of 13:45, 23 July 2012 by Bbrunner (talk | contribs) (Started adding smbta-documentation)

SMB Traffic Analyzer (SMBTA)

About SMBTA

What is SMB Traffic Analyzer

SMB Traffic Analyzer (from now on called SMBTA) is a software suite utilizing the Samba CIFS server to create statistics about the data traffic on a Samba network. SMBTA is made out of several components:

  • a module in the VFS (Virtual File System) Layer of Samba
  • a system daemon program
  • user oriented programs to visualize the statistics

SMBTA is utilizing a SQL storage to store data about the traffic caused on a network. By making this data available via SQL, SMBTA is helpful for many system administrators.

SMBTA also has real-time interfaces, making it possible to watch Samba traffic in real time on the console or interfacing with rrdtool to create a round robin database.

Having knowledge of SQL is not required. The user may also run the client programs to query the database. These are specialized to the organization of the database, and are taking away much of the work to translate often needed questions into SQL statements from the user. They are also running networked, and can be run on a complete different system where the SQL storage exists.

The basic idea and main concept of SMBTA was born on the SambaXP conference in Göttingen, Germany, in 2007. At the SNIA conference 2008, I was able to present a prototype of the idea to the Samba team. The code for the module got immediately accepted and was introduced with Samba 3.2.0 to Samba users.

This document is refering to the version 2.0 of the SMB Traffic Analyzer VFS module.

Please check SMB Traffic Analyzer’s home page for newer versions of the software or this document.

Why using SMB Traffic Analyzer

Samba services may run on special hardware, like RAID systems, or other specialized storage. With SMBTA, the user can create an overview of the network usage on the network. SMBTA can be used to answer questions like this:

  • Which of my services is the most used one?
  • At which time the usage of my service reaches peaks?
  • Which service almost never gets used?
  • Which users are the highest traffic generators?
  • What where the detailed last file activities on a share?
  • Which is the top used file by read and write access?

Isn’t SMBTA a tool to control employees?

Yes it can be used like that, but doesn’t need to be. To expose user names to a storage is even forbidden in many countries. Therefore, SMBTA allows for anonymization of any data linked to a person. There are many use cases for SMBTA, like determining and planning hardware for your services, or analyzing the network before switching to clustered Samba.

The basic concept of SMBTA

Samba servers configured with the VFS module send meta data to smbtad, which provides database access to the smbtatools programs

The Samba CIFS server features a Virtual File System Layer (VFS), that allows modules to be linked into the layer, replacing or enhancing the functions of the VFS. Modules can be linked into and behave completely transparent. SMBTA makes use of this functionality by adding a module to Samba’s VFS. In the VFS layer, the module is collecting data from prominent VFS functions like write, read, close etc.

The collected data may be be encrypted, and send over the network to a receiver program. This daemon program, smbtad, is then building a SQL storage from the data. Besides of the generated SQL storage, client programs can be used to query the storage in a specialized way over the network.

Where to get SMBTA

The SMBTA VFS module is called vfs_smb_traffic_analyzer.so, and is packaged with Samba version greater or equal than 3.2.0.

Important: The current SMBTA module shipping with Samba (any version older than 3.6.0pre1) is only implementing version 1 of the SMBTA protocol, and this document is only refering to version 2 of the protocol. The implementation of vfs_smb_traffic_analyzer.so protocol version 2 can be found in the current master git source tree of Samba. It is expected to be shipped with Samba 3.6.0.
Tip: For the openSUSE Linux distribution, SMBTAv2 has been backported to current Samba releases, from 3.5.0 and newer. You can retrieve the SMBTAv2 VFS module from the openSUSE BuildService. Furthermore, RPM packages for several openSUSE releases of the smbtatools and the smbtad component are available on the buildservice.

The other parts of SMBTA, the daemon program smbtad, as well as the client programs webSMBTA, smbtaquery and smbtamonitor are available in source code from the projects homepage at http://holger123.wordpress.com/smb-traffic-analyzer/.

Building SMBTA from the sources

The SMBTA tools smbtaquery and smbtamonitor are utilizing "cmake" as the primary build tool. The smbtad daemon program does too.

Before trying to build smbtad and smbtatools, check for the following requirements:

  • cmake ( required for both )
  • libsmbclient library ( required for smbtautils )
  • curses/ncurses library ( required for smbtautils )
  • libDBI ( required for both )

With cmake, you usually do an "out-of-source" build. That means that the source will not be build in the original source directory. Here is an example build. Suppose you’ve unpacked the smbtatools source tarball at (the same procedure can be followed to compile smbtad):

 /home/tom/smbtatools

You can now create your own directory where you want to build smbtatools. Like:

 mkdir /home/tom/smbtatools-build
 cd /home/tom/smbtatools-build

If you have any libraries or required components in other directories than the system paths, you should specify those by specifying environment variables before the cmake run:

 setenv CMAKE_INCLUDE_PATH $YOUR_PATH_TO_EXTERNAL_INCLUDES:$ANOTHER_DIRECTORY_WITH_EXTERNAL_INCLUDES
 setenv CMAKE_LIBRARY_PATH $YOUR_PATH_TO_LIBRARIES:$ANOTHER_DIRECTORY_WITH_LIBRARIES

Both smbtatools and smbtad require sqlite to run. If it isn’t installed on your system, or isn’t available in a version new enough, both packages will build an included sqlite variant and link to this.

You are now in the build directory. Just run cmake from here:

 cmake ../smbtatools/

If you want the later "make install" to install control files for SMBTA into /etc/init.d, set this cmake cache value:

 cmake -D CMAKE_INSTALL_RCFILE:string=yes ../smbtad/

And cmake will check the requirements of smbtatools/smbtad against your system. If it complains, it will return human readable messages about what is missing. After a successful cmake run, you can just do "make" and "make install" as "usual" in this directory.

After a succesful run of cmake, you can run "make" and "make install" ( and "make install" might require root privileges) as usual:

 make
 make install

If you want to install the package to a different prefix, you use:

 make DESTDIR=/$YOUR/$DESTINATION install

Setting up SMBTA

Upgrading from a former version

SMBTA provides an update path to convert databases used with former versions to the current one.

Important: If you are upgrading SMBTA from an older version, it is necessary to run this conversion process!

The smbtaquery program includes a command line option "-C --convert":

 smbtaquery -M $DB-DRIVER -N $DATABASE-NAME -S $DATABASE-USER -H $DATABASE-HOST  -C

This process is self explaining. Basically it presents a menu to choose the version of SMBTA you are coming from. Please consult the chapter about the smbtaquery program to read more about the arguments to the -M, -N, -S and -H options.

About the database

To run SMBTA, you’ll need to setup a database. Current valid databases are:

  • Postgresql, tested succesfully by the SMBTA development team.
  • sqlite3, tested by the SMBTA development team.
  • MySQL, untested by the SMBTA development team, but supported by libDBI. Feedback is highly welcome!

Furthermore, SMBTA will need the following information to access the database:

  • A name of a user that has access to the database
  • The users password
  • The hostname or IP adress of the system that is running the database
  • The name of the database to access.

Initial setup of the database

Important: This step is required if you install SMBTA for the first time!

If you are installing SMBTA for the first time, SMBTA assumes that the administrator has setup an empty database, with a user that has access to it. The smbtad component can then prepare an empty database on an initial run by using "smbtad -T", creating all required tables. See the chapter on smbtad for more about this.

Activating the VFS module on Samba Servers

In general, the Samba CIFS server is configured via smb.conf, it’s main configuration file. A typical service definition would look like this:

 [ExampleShare]
         path = /space
         read only = no

Defining the share "ExampleShare" as a CIFS service, refering to the path /space on the servers storage.

Internet socket

By extending the share definition with the commands to load the SMBTA VFS module, SMBTA logging for this service can be activated:

 [ExampleShare]
         path = /space
         read only = no
         vfs objects = smb_traffic_analyzer
         smb_traffic_analyzer:protocol_version = V2
         smb_traffic_analyzer:host = localhost
         smb_traffic_analyzer:port = 3490
Important: By default, SMBTA will use protocol version 1, to not break any existing installations. Only if the version parameter is given, it will be forced to use protocol version 2. This documentation is only refering to protocol version 2.

The host and port parameters define the hostname of the system that will be used as the target for the VFS modules data to send. It will try to connect to the port number given with the parameter port.

Unix domain socket

The module can also run using a unix domain socket, which is useful if you want to connect smbtad on the same system as the Samba server. In this case The share definition would look like this:

  [ExampleShare]
         path = /space
         read only = no
         vfs objects = smb_traffic_analyzer
         smb_traffic_analyzer:protocol_version = V2
         smb_traffic_analyzer:mode = unix_domain_socket

The mode parameter set to "unix_domain_socket" then will use a unix domain socket at /var/tmp/stadsocket.

Anonymization

Anonymization of user related information can be enabled in the VFS module. There are two modes of anonymization possible.

The first method is created by generating a hash number out of the username, and add the number to a given prefix in the configuration. Such, a user called "holger" might become "user123", if "user" was the given prefix.

 [ExampleShare]
         path = /space
         read only = no
         vfs_objects = smb_traffic_analyzer
         smb_traffic_analyzer:protocol_version = V2
         smb_traffic_analyzer:mode = unix_domain_socket
         # enable prefix based anonymization
         smb_traffic_analyzer:anonymize_prefix = user

The second method is totally anonymizing any users. It just replaces the username (and the SID) to a given prefix.

 [ExampleShare]
         path = /space
         read only = no
         vfs_objects = smb_traffic_analyzer
         smb_traffic_analyzer:protocol_version = V2
         smb_traffic_analyzer:mode = unix_domain_socket
         # enable total anonymization mapped to "user"
         smb_traffic_analyzer:anonymize_prefix = user
         smb_traffic_analyzer:total_anonymization = yes


Setting up smbtad

On the target host of the module, the program smbtad is running. It’s task is mainly to feed a SQL storage out of the data it receives from the module, maintain the storage and accept requests from clients.

 smbtad -i 3490 -p 3491

smbtad is a daemon program that can be run by any user. By default, it creates a directory '$HOME/.smbtad/' where it stores it’s database. The above call makes smbtad to listen to the VFS modules connections on port 3490, and listen to client connections on port 3491.

By default, smbtad creates a database as '$HOME/.smbtad/staddb'. If the database already exists, it will use the existing database.

The command line options of smbtad

Any option you give that smbtad doesn’t understand will lead to print out it’s list of processed commands. Such as if the user calls:

 smbtad --help

The following output will appear:

 SMB Traffic Analyzer daemon version 1.2.4
 License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
 (C) 2008-2011 Holger Hetterich <ozzy@metal-district.de>
 -S      --dbuser                Specifiy the user for the database.
 -H      --dbhost                Specifiy the host of the database.
 -P      --dbpassword            Specifiy to password to access the db.
 -M      --dbdriver              Specify the libDBI driver to use.
 -N      --dbname                Specify the name of the database.
 -i      --inet-port             Specifiy the port to be used.
                                 Default: 3490.
 -u      --unix-domain-socket    If this parameter is specified,
                                 a unix domain socket at
                                 /var/tmp/stadsocket will be
                                 used.
 -d      --debug-level           Specify the debug level (0-10).
                                 Default: 0.
 -q      --query-port            Port to be used for clients.
 -o      --interactive           Don't run as daemon.
                                 (Runs as daemon by default)
 -c      --config-file           Use configuration file given.
 -t --maintenance-timer <value>  specify the time intervall to
                                 to start the database
                                 maintenance routine. Format is
                                 HH:MM:SS
                                 Example: -m 00:30:00 will run
                                 the maintenance routine every
                                 half hour.
                                 Default: 01:00:00
 -m --maintenance-timer-config
      <value>                    specify a number of days
                                 and a time. Every database
                                 entry which is older than the
                                 the specified number of days
                                 will be deleted by the
                                 maintenance routine.
                                 Format is: DAYS, HH:MM:SS
                                 Default: 1,00:00:00
 -k --keyfile                    Keyfile for encryption to be used
                                 between module and smbtad.
 -p --precision                  Precision value for the build-in
                                 cache. Default is 5.
 -U --use-db                     Specify 0 or 1 as argument. If
                                 this is 0, no database handling
                                 will be done. Default is 1.
 -T --setup                      Do the initial database setup and exit.
  • -i --inet-port: Specifiy the internet socket port, that smbtad is using to listen for data from the VFS module. If the port number is not given, the Default setting 3490 will be used.
  • -d --debug-level: Specify the debug level when running smbtad. If you get smbtad crashing and want to produce a bug report, please run it with -d 10, which is the highest debug level. The default value for this setting is 0, only fatal errors will be reported in this case. The debug messages smbtad is producing will be consumed by syslog and thus appear in your system log.
  • -o --interactive: For debugging reasons, it can be useful to not run smbtad as a daemon program. When run with -o, smbtad will not become a daemon program. The default is to run in daemon mode.
  • -c --config-file: The user can provide a configuration file instead of providing the command line switches.
  • -q --query-port: This is the internet socket port to which clients can connect when those want to request real time information. In case this parameter is not given, it is set to 3491 (or in other words to the value of --i --inet-port + 1.
  • -t --maintenance-timer: To hinder the database from growing infinite, a maintenance process is included in smbtad, which will clean up the database after a certain rule (see -m --maintenance-timer-config). The -t option specifies the interval in hours, as to when the maintenance process should be run. For example, if you give 00:10:00, a maintenance process will be started every ten minutes.
  • -m --maintenance-timer-config: Here you can specify how the maintenance process should work. The format is DAYS, HH:MM:SS. For example, if you want to delete anything in the database that is older than one day, you would give 1,00:00:00 (which is the default). Would you like to delete anything in the database that is older than 2 weeks, then you would give: 14,00:00:00 as parameter. Having deleted everything that is older than 2 hours, would be 0,02:00:00 as parameter.
  • -u --unix-domain-socket: If this option is given, a unix domain socket under /var/tmp/stadsocket will be used for the connection to the VFS module. This is useful if smbtad is to be run on the same machine that runs the Samba server.
  • -n --unix-domain-socket-cl: If this option is given, a unix domain socket under /var/tmp/stadsocket_client will be used for the connection to clients such as smbtaquery. This is useful if smbtad is to be run on the same machine that runs the tools.

New options since 1.2.3:

  • -p --precision: As of version 1.2.3 of smbtad, the program will sum up similar VFS R/W entries in it’s cache and only store the sum of transferred bytes instead of every single VFS entry. It’s cache is insert-sort based and the summing up makes the resulting database much smaller than in former releases. Because of this interpolation, statistic results will get slightly unprecise. The -p/--precision argument allows the user to control the timespan in seconds, over which the cache is summing up similar VFS R/W entries. The default is set to 5 seconds. The more seconds the user is specifying, the smaller will the resulting database be, and the less precise will the results be, when queried through smbtaquery. To do real data-mining, this argument can be set to 0. In this mode, smbtad isn’t summing up any VFS entry, and stores them as is in the database. This is the behaviour of smtbad in former releases than 1.2.3.
  • -U --use_db: As of version 1.2.3 of smbtad, the database usage of smbtad can be completely shut down. This is useful when SMBTA is only to be used with rrddriver or smbtamonitor, which only rely on real time data. The option requires an integer argument that is either 0 or 1. If it is 1, which is the default, the database will be handled, if it’s 0, only real time data is supported. Please note, that under --use-db=0 circumstances, some monitor functionality such as the TOTAL monitor, which is supposed to transfer the TOTAL sum of transfered bytes of an object will have a different behaviour. As the database is not available, it cannot query for the initial sum and starts with 0. Also, the real time tools rrddriver and smbtamonitor are doing an identification procedure by default, which must be shut down by using the --I/--identify switch with the tools. If you want to use SMBTA for rrddriver or smbtamonitor only, it is recommended to set this option to 0.

New options for database-settings with libDBI since 1.2.4:

  • -S --dbuser: Specifies the user of the database, and must be a valid user of the database.
  • -H --dbhost: Specifies the host name of the system where the database runs.
  • -P --dbpassword: Specifies the password of the user (given with -S --dbuser to access the database.
  • -M --dbdriver: Specifies the driver to use for the database connection.
    • Database drivers are:
      • "pgsql" for Postgresql.
      • "mysql" for MySQL (untested).
      • "sqlite3" for Sqlite3 (untested).
  • -N --dbname: Specifies the name of the database to be used for the database connection.
  • -T --setup: As of version 1.2.4 of smbtad, the database needs to be setup by the -T --setup switch. This will access the database, and create the initial tables and structure of the database. smbtad will not daemonize after this call, just return normally.
  • -I --ip: This command line switch specifies the interface or network address to bind smbtad to when in network operation. It can either be a IPv4 address, a IPv6 address, or a fully qualified hostname which will be resolved by smbtad. If this option is not given, it’ll default to "localhost".