Centralized authentication using OpenLDAP

This guide introduces the basics of LDAP and shows you how to setup OpenLDAP for authentication purposes between a group of computers.

What is LDAP?
LDAP stands for Lightweight Directory Access Protocol. Based on X.500 it encompasses most of its primary functions, but lacks the more esoteric functions that X.500 has. Now what is this X.500 and why is there an LDAP?

X.500 is a model for Directory Services in the OSI concept. It contains namespace definitions and the protocols for querying and updating the directory. However, X.500 has been found to be overkill in many situations. Enter LDAP. Like X.500 it provides a data/namespace model for the directory and a protocol. However, LDAP is designed to run directly over the TCP/IP stack. See LDAP as a slim-down version of X.500.

I don't get it. What is a directory?
A directory is a specialized database designed for frequent queries but infrequent updates. Unlike general databases they don't contain transaction support or roll-back functionality. Directories are easily replicated to increase availability and reliability. When directories are replicated, temporary inconsistencies are allowed as long as they get synchronised eventually.

How is information structured?
All information inside a directory is structured hierarchically. Even more, if you want to enter data inside a directory, the directory must know how to store this data inside a tree. Lets take a look at a fictional company and an Internet-like tree:

Since you don't feed data to the database in this ascii-art like manner, every node of such a tree must be defined. To name such nodes, LDAP uses a naming scheme. Most LDAP distributions (including OpenLDAP) already contain quite a number of predefined (and general approved) schemas, such as the inetOrgPerson, or a frequently used schema to define users which Unix/Linux boxes can use, called posixAccount. Note there are GUI web based tools to make managing LDAP painless: see Working with OpenLDAP for an non-exhaustive list.

Interested users are encouraged to read the OpenLDAP Admin Guide.

So... What can it be used for?
LDAP can be used for various things. This document focuses on centralised user management, keeping all user accounts in a single LDAP location (which doesn't mean that it's housed on a single server, LDAP supports high availability and redundancy), yet other goals can be achieved using LDAP as well.


 * Public Key Infrastructure


 * Shared Calendar


 * Shared Addressbook


 * Storage for DHCP, DNS, ...


 * System Class Configuration Directives (keeping track of several server configurations)


 * Centralised Authentication (PosixAccount)



Common notes
The domain genfic.org is an example in this guide. You will of course want to change this. However, make sure that the top node is an official top level domain (net, com, cc, be, ...).

Let's first emerge OpenLDAP. Ensure the USE flags berkdb, crypt, gnutls, ipv6, sasl, ssl, syslog, -minimal and tcpd are used.

OpenLDAP has a main user called "rootdn" (Root Distinguished Name), which is hardcoded in the application. Unlike the classic Unix root user, the rootdn user still needs to be assigned with proper permissions. The rootdn user may be used only in the context of the configuration, but it can also be used in the directory definition. In that case a user can authenticate himself as rootdn with either the configuration used password and the tree (directory-based) password.

User passwords (regardless if it is for rootdn users or others) for verification purposes can be stored as cleartext or hashed. Multiple different hash algorithms are available, but usage of weak algorythms (up to MD5) is not recommended. SHA is currently considered sufficiently cryptographically secure.

In the below command, a hashed value is created for a given password; the result of this command can be used in the configuration file, or in the internal directory definition of a user:

For now (stable OpenLDAP 2.4.44) you probably should not need to use simple (username/password) mech, preferring SASL. In this case your ident description is provided by  utility.

Legacy configuration (flat config slapd.conf)
Now edit the LDAP Server configuration in. The provided is from the original OpenLDAP source. Below is a sample configuration file one can use to replace it with to get things started.

For a more detailed analysis of the configuration file, we suggest that you work through the OpenLDAP Administrator's Guide, although may be enough.

If it does not start, the first thing you must do is to check the config file. You can do it with the following command.

Vary the debug level (the "-d 1" above) for more info. If all goes well you will see config file testing succeeded. If there's an error,  will list the line number to which it applies (of the  file).

By default  write logs to   facility.

Note that since version 2.4.23, OpenLDAP default finally moved from traditional flat config files to OLC (OnLineConfiguration, also known through its   structure) as default configuration method. One of benefits of using OLC is that the dynamic backend doesn't require restart of server after updating the configuration. Existing users can migrate to the new configuration method by invoking  setting both -f and -F options. Traditionally OLC is stored in ldif back-end (which keep benefits of human-readability) in the  directory. In Gentoo it is not required to convert the configuration yet, but support for the currently documented approach will be removed in the future.

Migration from slapd.conf to OLC
If you want to be able to change OpenLDAP server's configuration, you must define at least write (or normally manage) access to.

The example below shows how to grant manage access to OLC (cn=config database) to the system administrator (root user) by adding the proper lines at the end of the file:

utility, invoked with both -f and -F options will convert into ready to use (just take care about proper permissions) configuration directory.

Running this command will transfer and translate the configuration. After that you are expected to update the configuration using specially prepared ldif files. And only if you aren't enough familiar with them, you can first edit and after that re-translate the  into. Don't forget to check the directory's permissions.

For more instructions read the in-line comments of the generated files.

The below line will enable the configuration method.

Finally, create the structure:

Here you need to load database dump:

When using legacy and planned to be removed BDB-backend it will be converted into numerous files. See for details.

When using modern (default since 2.4.40) MDB-backend all data is stored into two files: and. See for details.

Start slapd:

Initial setup with OLC
Initial config is shipped as standard LDAP database dump.

It could be loaded (and only loaded, unlike ordinar LDAP database) by  utility:

If you expect right to change configuration database, you must provide proper permissions (example giving these permissions for root system user without any additional confirmations from my production directory):

See for details.

Using OLC you must NOT manually edit it's configs, so, probably it's impossible to get broken config directory.

But if you want, you may check it's consistancy with the following command:

Server management with OLC
Some examples of updates on the OLC-style configuration are mentioned below.

For instance, to change the location of the OLC configuration directory (needed after translation config file to config directory):

To change the log level used by the OpenLDAP instance:

To add the syncprov overlay:

In order to apply the changes, run the following command:

OpenLDAP logging
OpenLDAP produce numerous logs, not very clear to read, but necessary for debug. So, we must describe this part.

It's rather important to repeat: if you want to have readable system logs, you must write slapd's log messages into dedicated log file, or disable it's logging at all.

To prevent DoS syslog daemons often limit amount of accepted messages from process in a period of time. For example with rsyslog:

To be continued…

Access management (ACLs)
The next mandatory question is handling permissions on database's.



access to [ by [ ] [ ] ]+

For details about exact privilege setting see manual pages and official docs.

Config file
ACLs are parsed, starting from the end of file, overriding previously set permissions.

For example:

Means that first ACL allows everybody to read the tree.

Second ACL overrides permissions of the first ACL allowing access to subschemas for authenticated users.

Third ACL overrides permissions of the first ACL limiting access to userPassword attribute, allowing self and replication user read it and anonymous auth. Remember, that rootdn could read and write everything.

Config directory
Access control is managed by enumerated olcAccess directives, similiarly starting from maximum number to the first (zero).

Example ldifs to add permissions, same with config, quoted above:

Insert ACL example:

Numbers of all existing ACLs, starting from inserted one, grew on 1. 0 became 1, 1 becvae 2 and so long.

Delete ACL example:

SASL authentication
Also mandatory for description, but not even planned yet.

Most common mech is EXTERNAL:

Specifies the  Security  Strength  Factor  (SSF)  to  be given local LDAP sessions, such as those to the ldapi:// listener. For a description of SSF values, see sasl-secprops's minssf option description. The default is 71.

Primary develop/debug util —.

Configuring the OpenLDAP Client tools
Edit the LDAP Client configuration file. This file is read by ldapsearch and other ldap command line tools.

You can test the running server with the following command:

If you receive an error, try adding  to increase the verbosity and solve the issue you have.

Client Configuration for Centralised Authentication
There are numerous methods/tools that can be used for remote authentication. Some distributions also have their own easy to use configuration tool. Below there are some in no particular order. It is possible to combine local users and centrally authorized accounts at the same time. This is important because, for instance, if the LDAP server cannot be accessed one can still login as root.


 * SSSD (Single Sign-on Services Daemon). Its primary function is to provide access to identity and authentication remote resource through a common framework that can provide caching and offline support to the system. It provides PAM and NSS modules, and in the future will support D-Bus interfaces for extended user information. It also provides a better database to store local users as well as extended user data.


 * Use  to login to the LDAP server and authenticate. Passwords are not sent over the network in clear text.


 * NSLCD (Name Service Look up Daemon). Similar to SSSD, but older.


 * NSS (Name Service Switch) using the traditional  module to fetch password hashes over the network. To permit users to update their password this has to be combined with the   method.

The first two are demonstrated below with the minimum necessary configuration options to get working.

Client PAM configuration SSSD Method
Here is the more direct method. The three files that are required to be edited are mentioned below.

Add sss to the end as shown below to enable the lookup to be handed to the sssd system service. Once you have finished editing start the sssd daemon.

The last file is the most critical. Open an extra root terminal as a fallback before editing this. The lines that end with  have been added to enable remote authentication. Note the use of to support creating the user home directories.

Now try logging in from another box.

Client PAM configuration the pam_ldap Module Method
First, we will configure PAM to allow LDAP authorization. Install so that PAM supports LDAP authorization, and  so that your system can cope with LDAP servers for additional information (used by ).

The last file is the most critical. Open a few extra root terminals as a backup before editing this. The lines that end with  have been added to enable remote authentication.

Now change to read:

Next, copy over the (OpenLDAP) file from the server to the client so the clients are aware of the LDAP environment:

Finally, configure your clients so that they check the LDAP for system accounts:

If you noticed one of the lines you pasted into your was commented out (the   line): you don't need it unless you want to change a user's password as superuser. In this case you need to echo the root password to in plaintext. This is DANGEROUS and should be chmoded to 600. What you might want to do is keep that file blank and when you need to change someone's password that's both in the LDAP and, put the pass in there for 10 seconds while changing the users password and remove it when done.

Migrate existing data to LDAP
Configuring OpenLDAP for centralized administration and management of common Linux/Unix items isn't easy, but thanks to some tools and scripts available on the Internet, migrating a system from a single-system administrative point-of-view towards an OpenLDAP-based, centralized managed system isn't hard either.

Go to http://www.padl.com/OSS/MigrationTools.html and fetch the scripts there. You'll need the migration tools and the script.

Next, extract the tools and copy the script inside the extracted location:

The next step now is to migrate the information of your system to OpenLDAP. The script will do this for you, after you have provided it with the information regarding your LDAP structure and environment.

At the time of writing, the tools require the following input:

The tool will also ask you which accounts and settings you want to migrate.

High availability
To setup replication of changes across multiple LDAP systems. Replication within OpenLDAP is, in this guide, set up using a specific replication account which has read rights on the primary LDAP server and which pulls in changes from the primary LDAP server to the secondary.

This setup is then mirrored, allowing the secondary LDAP server to act as a primary. Thanks to OpenLDAP's internal structure, changes are not re-applied if they are already in the LDAP structure.

Setting Up Replication
To setup replication, first setup a second OpenLDAP server, similarly as above. However take care that, in the configuration file:


 * The sync replication provider is pointing to the other system


 * The serverID of each OpenLDAP system is different

Next, create the synchronisation account. We will create an LDIF file (the format used as data input for LDAP servers) and add it to each LDAP server:

OpenLDAP permissions
If we take a look at you'll see that you can specify the ACLs (permissions if you like) of what data users can read and/or write:

This gives you access to everything a user should be able to change. If it's your information, then you got write access to it; if it's another user their information then you can read it; anonymous people can send a login/pass to get logged in. There are four levels, ranking them from lowest to greatest:.

The next ACL is a bit more secure as it blocks normal users to read other people their shadowed password:

This example gives root and John access to read/write/search for everything in the the tree below. This also lets users change their own 's. As for the ending statement everyone else just has a search ability meaning they can fill in a search filter, but can't read the search results. Now you can have multiple ACLs but the rule of the thumb is it processes from bottom up, so your toplevel should be the most restrictive one.

Maintaining the directory
You can start using the directory to authenticate users in apache/proftpd/qmail/samba. You can manage it with LAM (Ldap Account Manager), phpldapadmin, diradm, jxplorer, or lat, which provide easy management interfaces.

Acknowledgements
We would like to thank Matt Heler for lending us his box for the purpose of this guide. Thanks also go to the cool guys in #ldap @ irc.freenode.net