As of PowerDNS Authoritative Server 4.0.0, the LDAP backend is fully supported.
The original author for this module is Norbert Sendetzky. This page is based on the content from his LDAPbackend wiki section as copied in February 2016, and edited from there.
Host names and the MNAME of a SOA records are NEVER terminated with a '.' in PowerDNS storage! If a trailing '.' is present it will inevitably cause problems, problems that may be hard to debug.
The LDAP backend enables PowerDNS to retrieve DNS information from any standard compliant LDAP server. This is extremely handy if information about hosts is already stored in an LDAP tree.
The schema is based on the 'uninett' dnszone schema, with a few types added by number as designed in that schema:
The LDAP dnsdomain2 schema contains the additional object descriptions which are required by the LDAP server to check the validity of entries when they are added. Please consult the documentation of the LDAP server to find out how to add this schema to the server.
The LDAP backend can be compiled by adding
ldap to either the
When using packages, the
pdns-backend-ldap package should be installed.
There are a few options through the LDAP DNS backend can be configured. Add them to the
To launch the ldap backend:
ldap://127.0.0.1:389/") : The values assigned to this parameter can be LDAP URIs (e.g.
ldaps://127.0.0.1/) describing the connection to the LDAP server. There can be multiple LDAP URIs specified for load balancing and high availability if they are separated by spaces. In case the used LDAP client library doesn't support LDAP URIs as connection parameter, use plain host names or IP addresses instead (both may optionally be followed by a colon and the port).
(default "no") : Use TLS encrypted connections to the LDAP server. This is only allowed if ldap-host is an
ldap:// URI or a host name / IP address.
(default: "5") : The number of seconds to wait for LDAP operations to complete.
(default: "5") : The number of attempts to make to re-establish a lost connection to the LDAP server.
(default: "simple") : How to authenticate to the LDAP server. Actually only two methods are supported: "simple", which uses the classical DN / password, or "gssapi", which requires a Kerberos keytab.
(default "") : Path to the object to authenticate against. Should only be used, if the LDAP server doesn't support anonymous binds and with the "simple" authmethod.
(default "") : Password for authentication against the object specified by ldap-binddn. Only used when "authmethod" is "simple".
(default: "") : Full path to the keytab file to use to authenticate. This is only used when "authmethod" is set to "gssapi". The keytab must, ideally, contain only one principal (or to put it otherwise, only the first principal found in the keytab will be used).
(default: "") : Full path to the Kerberos credential cache file to use. Actually only files are supported, and the "FILE:" prefix must not be set. The PowerDNS process must be able to write to this file and it must be the only one able to read it.
(default "") : The PowerDNS LDAP DNS backend searches below this path for objects containing the specified DNS information. The retrieval of attributes is limited to this subtree. This option must be set to the path according to the layout of your LDAP tree, e.g. ou=hosts,o=linuxnetworks,c=de is the DN to my objects containing the DNS information.
(default "simple") :
simple: Search the requested domain by comparing the associatedDomain attributes with the domain string in the question.
tree: Search entires by translating the domain string into a LDAP dn. Your LDAP tree must be designed in the same way as the DNS LDAP tree. The question for "myhost.linuxnetworks.de" would translate into "dc=myhost,dc=linuxnetworks,dc=de,ou=hosts=..." and the entry where this dn points to would be evaluated for dns records.
strict: Like simple, but generates PTR records from aRecords or aAAARecords. Using "strict", zone transfers for reverse zones are not possible.
(default "(:target:)" ) : LDAP filter for limiting AXFR results (zone transfers), e.g. (&(:target:)(active=yes)) for returning only entries whose attribute "active" is set to "yes".
(default "(:target:)" ) : LDAP filter for limiting IP or name lookups, e.g. (&(:target:)(active=yes)) for returning only entries whose attribute "active" is set to "yes".
First off adding master support to the LDAP backend needs a schema update. This is required as some metadata must be stored by PowerDNS, such as the last successful transfer to slaves. The new schema is available in schema/pdns-domaininfo.schema.
Once the schema is loaded the zones for which you want to be a master must be modified. The dn of the SOA record must have the object class
PdnsDomain, and thus the
PdnsDomainId attribute. This attribute is an integer that must be unique across all zones served by the backend. Furthermore the
PdnsDomainType must be equal to 'master' (lower case).
Here is an example LDIF of a zone that's ready for master operation (assuming the 'tree' style):
dn: dc=example,dc=com,ou=dns,dc=mycompany,dc=com objectClass: top objectClass: domainRelatedObject objectClass: dNSDomain2 objectClass: PdnsDomain dc: example associatedDomain: example.com nSRecord: ns1.example.com sOARecord: ns1.example.com. hostmaster.example.com. 2013031101 1800 600 1209600 600 mXRecord: 10 mx1.example.com PdnsDomainId: 1 PdnsDomainType: master PdnsDomainMaster: 192.168.0.2
You should have one attribute
PdnsDomainMaster per master serving this zone.
The DNS LDAP tree should be designed carefully to prevent mistakes, which are hard to correct afterwards. The best solution is to create a subtree for all host entries which will contain the DNS records. This can be done the simple way or in a tree style.
DN of a simple style example record (e.g. myhost.example.com):
DN of a tree style example record (e.g. myhost.test.example.com):
Each domain (or zone for BIND users) must include one object containing a SOA (Start Of Authority) record. This requirement applies to both forward and reverse zones. This object can also contain the attribute for a MX (Mail eXchange) and one or more NS (Name Server) records. These attributes allow one or more values, e.g. for a backup mail or name server:
dn:dc=example,ou=hosts,o=example,c=com objectclass:top objectclass:dcobject objectclass:dnsdomain objectclass:domainrelatedobject dc:example soarecord:ns.example.com email@example.com 1 1800 3600 86400 7200 nsrecord:ns.example.com mxrecord:10 mail.example.com mxrecord:20 mail2.example.com associateddomain:example.com
A simple mapping between name and IP address can be specified by an object containing an
arecord and an
dn:dc=server,dc=example,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain objectclass:domainrelatedobject dc:server arecord:10.1.0.1 arecord:192.168.0.1 associateddomain:server.example.com
Be aware of the fact that these examples work if
strict. For tree mode, all DNs will have to be modified according to the algorithm described in the section above.
Wild-card domains are possible by using the asterisk in the
associatedDomain value like it is used in the BIND zone files. The "dc" attribute can be set to any value in simple or strict mode - this doesn't matter.
dn:dc=any,dc=example,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain objectclass:domainrelatedobject dc:any arecord:192.168.0.1 associateddomain:*.example.com
In tree mode wild-card entries has to look like this instead:
dn:dc=*,dc=example,dc=de,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain objectclass:domainrelatedobject dc:* arecord:192.168.0.1 associateddomain:*.example.com
Aliases for an existing DNS object have to be defined in a separate LDAP object. One object should be create per alias (this is a must in tree mode) or add all aliases (as values of
associateddomain) to one object. The only thing which is not allowed is to create loops by using the same name in
associateddomain and in
dn:dc=server-aliases,dc=example,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain objectclass:domainrelatedobject dc:server-aliases cnamerecord:server.example.com associateddomain:proxy.example.com associateddomain:mail2.example.com associateddomain:ns.example.com
Aliases are optional. All alias domains can also be added to the associateddomain attribute. The only difference is that these additional domains aren't recognized as aliases anymore, but instead as a normal
dn:dc=server,dc=example,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain objectclass:domainrelatedobject dc:server arecord:10.1.0.1 associateddomain:server.example.com associateddomain:proxy.example.com associateddomain:mail2.example.com associateddomain:ns.example.com
Currently there are two options: Set
strict to have the code automatically derive PTR records from A and AAAA records in the tree. Or, in
tree modes, create additional objects explictly mapping each address to a PTR record.
simple modes, first create an object with an SOA record for the reverse-lookup zone(s) corresponding to the A and AAAA records that will be served:
dn:dc=1.10.in-addr.arpa,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain2 objectclass:domainrelatedobject dc:1.10.in-addr.arpa soarecord:ns.example.com firstname.lastname@example.org 1 1800 3600 86400 7200 nsrecord:ns.example.com associateddomain:1.10.in-addr.arpa
strict mode, no other objects are required -- reverse queries that correspond to an arecord or aaaarecord of an existing object will be automagically serviced using the associateddomain entry of that object.
simple mode, you must then create objects for each reverse mapping:
dn:dc=1.0,dc=1.10.in-addr.arpa,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain2 objectclass:domainrelatedobject dc:1.0 ptrrecord:server.example.com associateddomain:18.104.22.168.in-addr.arpa
Tree mode requires each component to be a dc element of its own:
dn:dc=1,dc=0,dc=1,dc=10,dc=in-addr,dc=arpa,ou=hosts,o=example,c=de objectclass:top objectclass:dnsdomain2 objectclass:domainrelatedobject dc:1 ptrrecord:server.example.com associateddomain:22.214.171.124.in-addr.arpa
To use this kind of record, add the dnsdomain2 schema to the configuration of ther LDAP server.
ldap-method=strict can not be used if zone transfers (AXFR) are needed to other name servers. Distributing zones can only be done directly via LDAP replication in this case, because for a full zone transfer the reverse records are missing.
There is a small utility in the PowerDNS distribution available called
../manpages/zone2ldap.1, which can convert zone files used by BIND to the ldif format. Ldif is a text file format containing information about LDAP objects and can be read by every standard compliant LDAP server.
zone2ldap needs the BIND
named.conf (usually located in /etc) as input and writes the dns record entries in ldif format to stdout:
zone2ldap --basedn=YOUR_BASE_DN \ --named-conf=PATH_TO_NAMED_CONF \ --resume > zones.ldif
Alternatively zone2ldap can be used to convert only single zone files instead all zones:
zone2ldap --basedn=YOUR_BASE_DN \ --zone-file=PATH_TO_ZONE_FILE \ --zone-name=NAME_OF_ZONE \ --resume > zone.ldif
its manpage <../manpages/zone2ldap.1> for a complete list of options.
When coming from the BIND LDAP sdb backend, the records can be kept in the LDAP tree also for the PowerDNS LDAP backend. The schemas both backends utilize is almost the same except for one important thing: Domains for PowerDNS are stored in the attribute "associatedDomain" whereas BIND stores them split in "relativeDomainName" and "zoneName".
There is a migration script which creates a file in LDIF format with the necessary LDAP updates including the "associatedDomain" and "dc" attributes. The utility is executed on the command line by:
./bind2pdns-ldap --host=HOSTNAME_OR_IP \ --basedn=YOUR_BASE_DN \ --binddn=ADMIN_DN > update.ldif
The parameter "host" and "basedn" are mandatory, "binddn" is optional. If "binddn" is given, the script will prompt for a password, otherwise an anonymous bind is executed. The updates in LDIF format are written to stdout and can be redirected to a file.
The script requires Perl and the Perl Net::LDAP module and can be downloaded here.
Updating the entries in the LDAP tree requires to make the dnsdomain2 schema known to the LDAP server. Unfortunately, both schemas (dnsdomain2 and dnszone) share the same record types and use the same OIDs so the LDAP server can't use both schemas at the same time. The solution is to add the dnsdomain2 schema and replace the dnszone schema by the dnszone-migrate schema. After restarting the LDAP server attributes from both schemas can be used and updating the objects in the LDAP tree using the LDIF file generated from
bind2pdns-ldap will work without errors.
The easiest way for migrating DNS records is to use the output of a zone transfer (AXFR). Save the output of the
dig program provided by BIND into a file and call
zone2ldap with the file name as option to the
--zone-file parameter. This will generate the appropriate ldif file, which can be imported into the LDAP tree. The bash script except below automates this:
DNSSERVER=127.0.0.1 DOMAINS="example.com 10.10.in-addr.arpa" for DOMAIN in $DOMAINS; do dig @$DNSSERVER $DOMAIN AXFR> $DOMAIN.zone; zone2ldap --zone-name=$DOMAIN --zone-file=$DOMAIN.zone> $DOMAIN.ldif; done
To improve performance, the LDAP server can maintain indices on certain attributes. This leads to much faster searches for these type of attributes.
The LDAP DNS backend mainly searches for values in
associatedDomain, so maintaining an index (pres,eq,sub) on this attribute is a big performance improvement:
ldap-method=strict is set, it uses the aRecord and aAAARecord attribute for reverse mapping of IP addresses to names. To maintain an index (pres,eq) on these attributes also improves performance of the LDAP server:
indexaAAARecord pres,eq indexaRecord pres,eq
All other attributes than associatedDomain, aRecord or aAAARecord are only read if the object matches the specified criteria. Thus, maintaining an index on these attributes is useless.
If the DNS-entries were added before adding these statements to
slapd.conf, the LDAP server will have to be stopped and
slapindex should be used on the command line. This will generate the indices for already existing attributes.
Converting the string in the dNSTTL attribute to an integer is a time consuming task. If no separate TTL value for each entry is requires, use the
setting-default-ttl parameter in
pdns.conf instead. This will gain a 7% improvement in performance for entries that aren't cached. A dNSTTL attribute can still be added to entries that should have a different TTL than the default TTL
The method of accessing the entries in the directory affects the performance too. By default, the "simple" method is used search for entries by using their associatedDomain attribute. Alternatively, the "tree" method can be used, whereby the search is done along the directory tree, e.g. "host.example.com" is translated into "dc=host,dc=example,dc=com,...". This requires the LDAP DNS subtree layout to be 1:1 to the DNS tree, this will gain an additional 7% performance improvement.
The LDAP tree must contain a separate subtree of PTR records (e.g. for 126.96.36.199.in-addr.arpa) and
ldap-method can't be set to "strict".
For automatically generated reverse IPv6 records the aAAARecord entries must follow two restrictions: They have to be fully expanded ("FFFF::1" is not allowed and it must be "FFFF:0:0:0:0:0:0:1" instead) and they must not contain leading zeros, e.g. an entry containing "002A" is incorrect - use "2A" without zeros instead. These restrictions are due to the fact that LDAP DNS AAAA entries are pure text and doesn't allow searching by wild-cards.
As soon as the LDAP server implementations begin to provide the features of the LDAP client update protocol (LCUP,
3928), it will be possible to support the DNS notification feature for the LDAP DNS backend in case a record in the LDAP directory was changed.
Support for more authentication methods would be handy. Anyone interested may contribute.