[Date Prev][Date Next]
FTree backend module (ITS#1499)
Full_Name: Alex Martin
Submission from: (NULL) (126.96.36.199)
The following files contain my FTree module + patches
Here is its README:
FTree slapd module
The FTree slapd module is designed to implement a LDAP DIT using a flexible
memory-based directory-like tree structure. Access to LDAP entries stored in
tree is protected by read-write mutexes in order to facilitate efficient
threaded operation. The tree structure is initiated from the slapd
file. A basic sample configuration could be:
The lines starting with "FtRoot" are internally converted into LDIF format and
used to initialize the root of the DIT. In this case the root would be set to
Once initialized, additional entries may added and deleted from backend using
normal LDAP operations.
Although a basic memory-based backend provides useful functionality, a key
design feature of the FTree module is the implementation of entries in the
memory-based structure as polymorphic types. The type information of entries is
defined by the value(s) of the ObjectClass attribute and the sub-classed
functionality is loaded into the server by means of additional modules (shared
libraries). When an entry is added to the server its type information is
determined and compared with those registered by any loaded modules. If it is
found to match it is associated with the corresponding function table. When a
type match is not found, basic default behaviour is assumed.
This method is used to implement the "special" LDAP entry types Alias,
and also DynamicObject ( rfc2589 ). The functionality to support these types
be loaded into the server by the addition of the lines:
to the slapd configuration file.
The concept of LDAP entries with overloaded functionality is expanded further
to allow additional user-defined types. The intention here is to provide an
API which can facilitate the addition of more modular functionality.
One use of this concept is the integration of external information sources
e.g. scripts, files and other LDAP servers into the DIT. At a basic level
this can be thought of as a flexible version of some of the existing
OpenLDAP backends i.e. back-shell, back-ldap. However, unlike the existing
backends, multiple entries corresponding to these external info sources
may be freely added and deleted from a server. The schema defining these
types are contained in the file
This functionality is loaded via the modules:
External is an abstract base class whose methods are implemented by the Exec
Ldap classes.Examples of their use are provided in the files exec-example.ldif
and ldap-example.ldif in the demos directory %SYSCONFDIR%/demos:
dn: in=loadaverage1, dc=localhost, dc=localdomain, ou=myOrg
This entry specifies the external info source is a command "loadaverage"
with arguments "in=loadaverage1,dc=localhost,dc=localdomain,ou=myOrg".
For security reasons, the path used to locate commands is defined by a class
variable which is set by the "execpath" variable in the configuation file.
The API used when forking the command is the same as that used in the standard
back-shell backend module.
A similar configuration specifying another LDAP server as an info source
dn: dc=localhost1, dc=localdomain, ou=myOrg
Here "ref" specifies the URL of the external server using the same
syntax as a referral.
An additional feature provided by the FTree backend is the ability to
setup a cache of an external info source. The basic functionality is
provided by abstract classes "Cache" and "ThreadedCache". "CacheSearches"
implements a simple non-threaded model where search results are cached.
Several threaded cache models are provided "CachePeriodically", where a thread
periodically updates the cache from its source and "CacheFromStream", where
the info source pushes updates into the cache. In addition there is a threaded
cache model "FlushPeriodically" in which the memory contents are flushed to a
backing store. All of this core functionality is loaded via:
"FlushPeriodically" may be used to provide a simple persistent store
for the contents of the server. A root configuration:
Maps the root of the server to the contents of the ldif file exec-example.ldif
residing in the directory %SYSCONFDIR%/demos. The ldif file
will be read when the server starts and the current in-memory contents of the
server will be flushed backed to this file every 60 secs.
Other Cache Examples are provided in the exec-example.ldif file:
dn: in=loadaverage2, dc=localhost, dc=localdomain, ou=myOrg
Specifies a simple cache in which the results of any search handled
by the "loadaverage" script are cached and assumed to be valid for 60 seconds.
dn: in=loadaverage3, dc=localhost, dc=localdomain, ou=myOrg
Specifies a threaded cache should be setup. A thread is started which
periodically updates the cache by running "loadaverage" every 60 seconds.
dn: in=loadaverage4, dc=localhost, dc=localdomain, ou=myOrg
Also specifies a threaded cache should be setup. However, in this case,
the command forked, "loadaverage-stream", is assumed to continue running
and stream updates back to the server. Individual update records are
indicated by a terminating "RESULT" string. If the command exits the
server thread will attempt to restart it after 60 seconds.
The same cache methods are also implemented when another LDAP server forms
the information source (see ldap-example.ldif ). In order to facilitate
the streaming of LDAP updates between servers the Persistent Search
(draft-ietf-ldapext-psearch-03.txt) draft standard has been adapted.
This may also be utilised by clients, and a simple example ( ldap_psearch )