masters { 192.168.4.12; };
Primitive load balancing can be achieved in DNS using multiple A records for
For example, if you have three WWW servers with network addresses of
10.0.0.1, 10.0.0.2 and 10.0.0.3, a set of records such as the following
means that clients will connect to each machine one third of the time:
NameTTL CLASS TYPE Resource Record (RR) Data
When a resolver queries for these records, BIND will rotate them and respond
to the query with the records in a different order. In the example above,
clients will randomly receive records in the order 1, 2, 3; 2, 3, 1; and 3,
1, 2. Most clients will use the first record returned and discard the rest.
For more detail on ordering responses, check the rrset-order substatement in
the options statement under RRset Ordering. This substatement is not
supported in BIND 9, and only the ordering scheme described above is
DNS Notify is a mechanism that allows master nameservers to notify their
slave servers of changes to a zone's data. In response to a NOTIFY from a
master server, the slave will check to see that its version of the zone is
the current version and, if not, initiate a transfer.
DNS Notify is fully documented in RFC 1996. See also the description of the
zone option also-notify under Zone Transfers . More information about notify
can be found under Boolean Options .
3.4 Nameserver Operations
3.4.1 Tools for Use With the Nameserver Daemon
There are several indispensable diagnostic, administrative and monitoring
tools available to the system administrator for controlling and debugging
the nameserver daemon. We describe several in this section
The domain information groper ( dig ) is a command line tool that can be
used to gather information from the Domain Name System servers. Dig has two
modes: simple interactive mode for a single query, and batch mode which
executes a query for each in a list of several query lines. All query
options are accessible from the command line.
dig [@server] domain [<query-type>] [<query-class>]
[+<query-option>] [-<dig-option>] [%comment]
The usual simple use of dig will take the form
dig @server domain query-type query-class
For more information and a list of available commands and options, see the
The host utility provides a simple DNS lookup using a command-line interface
for looking up Internet hostnames. By default, the utility converts between
host names and Internet addresses, but its functionality can be extended
host [-aCdlrTwv] [-c class] [-N ndots] [-t type]
[-W timeout] [-R retries] hostname [server]
nslookup is a program used to query Internet domain nameservers. nslookup
has two modes: interactive and non-interactive. Interactive mode allows the
user to query nameservers for information about various hosts and domains or
to print a list of hosts in a domain. Non-interactive mode is used to print
just the name and requested information for a host or domain.
nslookup [-option ...] [host-to-find | -[server]]
Interactive mode is entered when no arguments are given (the default
nameserver will be used) or when the first argument is a hyphen (`-') and
the second argument is the host name or Internet address of a nameserver.
Non-interactive mode is used when the name or Internet address of the host
to be looked up is given as the first argument. The optional second argument
specifies the host name or address of a nameserver.
The options listed under the "set" command (see the nslookup man page for
details) can be specified in the .nslookuprc file in the user's home
directory if they are listed one per line. Options can also be specified on
the command line if they precede the arguments and are prefixed with a
hyphen. For example, to change the default query type to host information,
and the initial time-out to 10 seconds, type:
nslookup -query=hinfo -timeout=10
For more information and a list of available commands and options, see the
Due to its arcane user interface and frequently inconsistent behavior, we do
not recommend the use of nslookup , and it is not installed by default when
installing BIND 9. Use dig instead.
3.4.1.2 Administrative Tools
Administrative tools play an integral part in the management of a server.
The remote name daemon control ( rndc ) program is a program that allows the
system administrator to control the operation of a nameserver. If you run
rndc without any options it will display a usage message.
rndc [-p port] [-m] server command [command ...]
For more information and a list of available commands and options, see the
------------------------------------------------------------------------
Section 4. Advanced Concepts
Dynamic update is the term used for the ability under certain specified
conditions to add, modify or delete records or RRsets in the master zone
files. Dynamic update is fully described in RFC 2136.
Dynamic update is enabled on a zone-by-zone basis, by including an
allow-update or update-policy clause in the zone statement.
Updating of secure zones (zones using DNSSEC) is modelled after the
simple-secure-update proposal, a work in progress in the DNS Extensions
working group of the IETF. (See
the DNS Extensions working group.) SIG and NXT records affected by updates
are automatically regenerated by the server using an online zone key. Update
authorization is based on transaction signatures and an explicit server
The zone files of dynamic zones must not be edited by hand. The zone file on
disk at any given time may not contain the latest changes performed by
dynamic update. The zone file is written to disk only periodically, and
changes that have occurred since the zone file was last written to disk are
stored only in the zone's journal ( .jnl ) file. BIND 9 currently does not
update the zone file when it exits as BIND 8 does, so editing the zone file
manually is unsafe even when the server has been shut down.
4.2 Incremental Zone Transfers (IXFR)
The incremental zone transfer (IXFR) protocol is a way for slave servers to
transfer only changed data, instead of having to transfer the entire zone.
The IXFR protocol is documented in RFC 1995. See the list of proposed
standards in Appendix C .
When acting as a master, BIND 9 supports IXFR for those zones where the
necessary change history information is available. These include master
zones maintained by dynamic update and slave zones whose data was obtained
by IXFR, but not manually maintained master zones nor slave zones obtained
by performing a full zone transfer (AXFR).
When acting as a slave, BIND 9 will attempt to use IXFR unless it is
explicitly disabled. For more information about disabling IXFR, see the
description of the request-ixfr clause of the server statement.
Setting up different views, or visibility, of DNS space to internal and
external resolvers is usually referred to as a Split DNS setup. There are
several reasons an organization would want to set up its DNS this way.
One common reason for setting up a DNS system this way is to hide "internal"
DNS information from "external" clients on the Internet. There is some
debate as to whether or not this is actually useful. Internal DNS
information leaks out in many ways (via email headers, for example) and most
savvy "attackers" can find the information they need using other means.
Another common reason for setting up a Split DNS system is to allow internal
networks that are behind filters or in RFC 1918 space (reserved IP space, as
documented in RFC 1918) to resolve DNS on the Internet. Split DNS can also
be used to allow mail from outside back in to the internal network.
Here is an example of a split DNS setup:
Let's say a company named Example, Inc. (
example.com) has several corporate
sites that have an internal network with reserved Internet Protocol (IP)
space and an external demilitarized zone (DMZ), or "outside" section of a
network, that is available to the public.
Example, Inc. wants its internal clients to be able to resolve external
hostnames and to exchange mail with people on the outside. The company also
wants its internal resolvers to have access to certain internal-only zones
that are not available at all outside of the internal network.
In order to accomplish this, the company will set up two sets of
nameservers. One set will be on the inside network (in the reserved IP
space) and the other set will be on bastion hosts, which are "proxy" hosts
that can talk to both sides of its network, in the DMZ.
The internal servers will be configured to forward all queries, except
nameservers must be configured to disallow all queries to these domains from
any external hosts, including the bastion hosts.
The external servers, which are on the bastion hosts, will be configured to
could include things such as the host records for public servers (
special MX records that contain wildcard (`*') records pointing to the
bastion hosts. This is needed because external mail servers do not have any
other way of looking up how to deliver mail to those internal hosts. With
the wildcard records, the mail will be delivered to the bastion host, which
can then forward it on to internal hosts.
Here's an example of a wildcard MX record:
Now that they accept mail on behalf of anything in the internal network, the
bastion hosts will need to know how to deliver mail to internal hosts. In
order for this to work properly, the resolvers on the bastion hosts will
need to be configured to point to the internal nameservers for DNS
Queries for internal hostnames will be answered by the internal servers, and
queries for external hostnames will be forwarded back out to the DNS servers
In order for all this to work properly, internal clients will need to be
configured to query only the internal nameservers for DNS queries. This
could also be enforced via selective filtering on the network.
If everything has been set properly, Example, Inc. 's internal clients will
* Look up any hostnames on the Internet.
* Exchange mail with internal AND external people.
Hosts on the Internet will be able to:
Here is an example configuration for the setup we just described above. Note
that this is only configuration information; for information on how to
configure your zone files, see the Sample Configurations .
Internal DNS server config:
acl internals { 172.16.72.0/24; 192.168.1.0/24; };
acl externals { bastion-ips-go-here; };
forwarders { bastion-ips-go-here; }; // forward to external servers
allow-transfer { none; }; // sample allow-transfer (no one)
allow-query { internal; externals; }; // restrict query access
allow-recursion { internals; }; // restrict recursion
forwarders { }; // do normal iterative
// resolution (do not forward)
allow-query { internals; externals; };
allow-transfer { internals; };
masters { 172.16.72.3; };
allow-query { internals; externals; };
allow-transfer { internals; };
allow-query { internals; };
allow-transfer { internals; }
masters { 172.16.72.3; };
allow-query { internals };
allow-transfer { internals; }
External (bastion host) DNS server config:
acl internals { 172.16.72.0/24; 192.168.1.0/24; };
allow-transfer { none; }; // sample allow-transfer (no one)
allow-query { internals; externals; }; // restrict query access
allow-recursion { internals; externals; }; // restrict recursion
allow-transfer { internals; externals; };
masters { another_bastion_host_maybe; };
allow-transfer { internals; externals; }
In the
resolv.conf (or equivalent) on the bastion host(s):
This is a short guide to setting up Transaction SIGnatures (TSIG) based
transaction security in BIND. It describes changes to the configuration file
as well as what changes are required for different features, including the
process of creating transaction keys and using transaction signatures with
BIND primarily supports TSIG for server to server communication. This
includes zone transfer, notify, and recursive query messages. Resolvers
based on newer versions of BIND 8 have limited support for TSIG.
TSIG might be most useful for dynamic update. A primary server for a dynamic
zone should use access control to control updates, but IP-based access
control is insufficient. Key-based access control is far superior. See RFC
2845 in the Proposed Standards section of the Appendix. The nsupdate program
that is shipped with BIND 8 supports TSIG via the " -k " command line
4.4.1 Generate Shared Keys for Each Pair of Hosts
A shared secret is generated to be shared between host1 and host2 . An
arbitrary key name is chosen: "host1-host2.". The key name must be the same
4.4.1.1 Automatic Generation
The following command will generate a 128 bit (16 byte) HMAC-MD5 key as
described above. Longer keys are better, but shorter keys are easier to
read. Note that the maximum key length is 512 bits; keys longer than that
will be digested with MD5 to produce a 128 bit key.
dnssec-keygen -a hmac-md5 -b 128 -n HOST host1-host2.
The key is in the file Khost1-host2.+157+
00000.private . Nothing directly
uses this file, but the base-64 encoded string following " Key :" can be
extracted from the file and used as a shared secret:
The string "
La/E5CjG9O+os1jq0a2jdA== " can be used as the shared secret.
4.4.1.2 Manual Generation
The shared secret is simply a random sequence of bits, encoded in base-64.
Most ASCII strings are valid base-64 strings (assuming the length is a
multiple of 4 and only valid characters are used), so the shared secret can
Also, a known string can be run through mmencode or a similar program to
generate base-64 encoded data.
4.4.2 Copying the Shared Secret to Both Machines
This is beyond the scope of DNS. A secure transport mechanism should be
used. This could be secure FTP, ssh, telephone, etc.
4.4.3 Informing the Servers of the Key's Existence
Imagine host1 and host 2 are both servers. The following is added to each
The algorithm, hmac-md5, is the only one supported by BIND. The secret is
the one generated above. Since this is a secret, it is recommended that
either
named.conf be non-world readable, or the key directive be added to a
non-world readable file that is included by
named.conf .
At this point, the key is recognized. This means that if the server receives
a message signed by this key, it can verify the signature. If the signature
succeeds, the response is signed by the same key.
4.4.4 Instructing the Server to Use the Key
Since keys are shared between two hosts only, the server must be told when
keys are to be used. The following is added to the
named.conf file for host1
, if the IP address of host2 is 10.1.2.3:
Multiple keys may be present, but only the first is used. This directive
does not contain any secrets, so it may be in a world-readable file.
If host1 sends a message that is a response to that address, the message
will be signed with the specified key. host1 will expect any responses to
signed messages to be signed with the same key.
A similar statement must be present in host2 's configuration file (with
host1 's address) for host2 to sign non-response messages to host1 .
4.4.5 TSIG Key Based Access Control
BIND allows IP addresses and ranges to be specified in ACL definitions and
allow-{ query | transfer | update } directives. This has been extended to
allow TSIG keys also. The above key would be denoted key host1-host2.
An example of an allow-update directive would be:
allow-update { key host1-host2. ;};
This allows dynamic updates to succeed only if the request was signed by a
key named " host1-host2. ".
The more powerful update-policy statement is described Dynamic Update
The processing of TSIG signed messages can result in several errors. If a
signed message is sent to a non-TSIG aware server, a FORMERR will be
returned, since the server will not understand the record. This is a result
of misconfiguration, since the server must be explicitly configured to send
a TSIG signed message to a specific server.
If a TSIG aware server receives a message signed by an unknown key, the
response will be unsigned with the TSIG extended error code set to BADKEY.
If a TSIG aware server receives a message with a signature that does not
validate, the response will be unsigned with the TSIG extended error code
set to BADSIG. If a TSIG aware server receives a message with a time outside
of the allowed range, the response will be signed with the TSIG extended
error code set to BADTIME, and the time values will be adjusted so that the
response can be successfully verified. In any of these cases, the message's
TKEY is a mechanism for automatically generating a shared secret between two
hosts. There are several "modes" of TKEY that specify how the key is
generated or assigned. BIND implements only one of these modes, the
Diffie-Hellman key exchange. Both hosts are required to have a
Diffie-Hellman KEY record (although this record is not required to be
present in a zone). The TKEY process must use signed messages, signed either
by TSIG or SIG(0). The result of TKEY is a shared secret that can be used to
sign messages with TSIG. TKEY can also be used to delete shared secrets that
it had previously generated.
The TKEY process is initiated by a client or server by sending a signed TKEY
query (including any appropriate KEYs) to a TKEY-aware server. The server
response, if it indicates success, will contain a TKEY record and any
appropriate keys. After this exchange, both participants have enough
information to determine the shared secret; the exact process depends on the
TKEY mode. When using the Diffie-Hellman TKEY mode, Diffie-Hellman keys are
exchanged, and the shared secret is derived by both participants.
BIND 9 partially supports DNSSEC SIG(0) transaction signatures as specified
in RFC 2535. SIG(0) uses
public/private keys to authenticate messages.
Access control is performed in the same manner as TSIG keys; privileges can
be granted or denied based on the key name.
When a SIG(0) signed message is received, it will only be verified if the
key is known and trusted by the server; the server will not attempt to
locate
and/or validate the key.
BIND 9 does not ship with any tools that generate SIG(0) signed messages.
Cryptographic authentication of DNS information is possible through the DNS
Security ( DNSSEC ) extensions, defined in RFC 2535. This section describes
the creation and use of DNSSEC signed zones.
In order to set up a DNSSEC secure zone, there are a series of steps which
must be followed. BIND 9 ships with several tools that are used in this
process, which are explained in more detail below. In all cases, the " -h "
option prints a full list of parameters.
There must also be communication with the administrators of the parent
and/or child zone to transmit keys and signatures. A zone's security status
must be indicated by the parent zone for a DNSSEC capable resolver to trust
For other servers to trust data in this zone, they must either be statically
configured with this zone's zone key or the zone key of another zone above
this one in the DNS tree.
The dnssec-keygen program is used to generate keys.
A secure zone must contain one or more zone keys. The zone keys will sign
all other records in the zone, as well as the zone keys of any secure
delegated zones. Zone keys must have the same name as the zone, a name type
of ZONE , and must be usable for authentication. It is recommended that zone
keys be mandatory to implement a cryptographic algorithm; currently the only
key mandatory to implement an algorithm is DSA.
The following command will generate a 768 bit DSA key for the
child.exampleidentifier). The key file names contain the key name (
child.example. ),
algorithm (3 is DSA, 1 is RSA, etc.), and the key identifier (12345 in this
case). The private key (in the .private file) is used to generate
signatures, and the public key (in the .key file) is used for signature
To generate another key with the same properties (but with a different key
identifier), repeat the above command.
The public keys should be inserted into the zone file with $INCLUDE
statements, including the .key files.
The dnssec-makekeyset program is used to create a key set from one or more
Once the zone keys have been generated, a key set must be built for
transmission to the administrator of the parent zone, so that the parent
zone can sign the keys with its own zone key and correctly indicate the
security status of this zone. When building a key set, the list of keys to
be included and the TTL of the set must be specified, and the desired
signature validity period of the parent's signature may also be specified.
The list of keys to be inserted into the key set may also included non-zone
keys present at the apex. dnssec-makekeyset may also be used at non-apex
The following command generates a key set containing the above key and
another key similarly generated, with a TTL of 3600 and a signature validity
period of 10 days starting from now.
transmitted to the parent to be signed. It includes the keys, as well as
signatures over the key set generated by the zone keys themselves, which are
used to prove ownership of the private keys and encode the desired validity
4.7.3 Signing the Child's Keyset
The dnssec-signkey program is used to sign one child's keyset.
If the
child.example zone has any delegations which are secure, for example,
files for each secure subzone. These keys must be signed by this zone's zone
The following command signs the child's key set with the zone keys:
should be both transmitted back to the child and retained. It includes all
keys (the child's keys) from the keyset file and signatures generated by
The dnssec-signzone program is used to sign a zone.
Any signedkey files corresponding to secure subzones should be present, as
well as a signedkey file for this zone generated by the parent (if there is
one). The zone signer will generate NXT and SIG records for the zone, as
well as incorporate the zone key signature from the parent and indicate the
security status at all delegation points.
The following command signs the zone, assuming it is in a file called
private key are used to generate signatures.
referenced by
named.conf as the input file for the zone.
4.7.5 Configuring Servers
Unlike in BIND 8, data is not verified on load in BIND 9, so zone keys for
authoritative zones do not need to be specified in the configuration file.
The public key for any security root must be present in the configuration
trusted-keys statement, as described later in this document.
4.8 IPv6 Support in BIND 9
BIND 9 fully supports all currently defined forms of IPv6 name to address
and address to name lookups. It will also use IPv6 addresses to make queries
when running on an IPv6 capable system.
For forward lookups, BIND 9 supports both A6 and AAAA records. The of AAAA
records is deprecated, but it is still useful for hosts to have both AAAA
and A6 records to maintain backward compatibility with installations where
AAAA records are still used. In fact, the stub resolvers currently shipped
with most operating system support only AAAA lookups, because following A6
chains is much harder than doing A or AAAA lookups.
For IPv6 reverse lookups, BIND 9 supports the new "bitstring" format used in
the
ip6.arpa domain, as well as the older, deprecated "nibble" format used
BIND 9 includes a new lightweight resolver library and resolver daemon which
new applications may choose to use to avoid the complexities of A6 chain
following and bitstring labels. See The BIND 9 Lightweight Resolver for more
4.8.1 Address Lookups Using AAAA Records
The AAAA record is a parallel to the IPv4 A record. It specifies the entire
address in a single record. For example,
host 1h IN AAAA 3ffe:8050:201:1860:42::1
While their use is deprecated, they are useful to support older IPv6
applications. They should not be added where they are not absolutely
4.8.2 Address Lookups Using A6 Records
The A6 record is more flexible than the AAAA record, and is therefore more
complicated. The A6 record can be used to form a chain of A6 records, each
specifying part of the IPv6 address. It can also be used to specify the
entire record as well. For example, this record supplies the same data as
the AAAA record in the previous example:
host 1h IN A6 0 3ffe:8050:201:1860:42::1
A6 records are designed to allow network renumbering. This works when an A6
record only specifies the part of the address space the domain owner
controls. For example, a host may be at a company named "company." It has
two ISPs which provide IPv6 address space for it. These two ISPs fully
specify the IPv6 prefix they supply.
In the company's address space:
company 1h IN A6 0 3ffe:8050:201:1860::
company 1h IN A6 0 1234:5678:90ab:fffa::
caching name server) will find two partial A6 records, and will use the
additional name to find the remainder of the data.
4.8.2.2 A6 Records for DNS Servers
When an A6 record specifies the address of a name server, it should use the
full address rather than specifying a partial address. For example:
ns0 4h IN A6 0 3ffe:8050:201:1860:42::1
It is recommended that IPv4-in-IPv6 mapped addresses not be used. If a host
has an IPv4 address, use an A record, not an A6, with ::ffff:192.168.42.1 as
4.8.3 Address to Name Lookups Using Nibble Format
While the use of nibble format to look up names is deprecated, it is
supported for backwards compatiblity with existing IPv6 applications.
When looking up an address in nibble format, the address components are
simply reversed, just as in IPv4, and
ip6.int. is appended to the resulting
name. For example, the following would provide reverse name lookup for a
host with address 3ffe:8050:201:1860:42::1 .
4.8.4 Address to Name Lookups Using Bitstring Format
Bitstring labels can start and end on any bit boundary, rather than on a
multiple of 4 bits as in the nibble format. They also use
ip6.arpa rather
To replicate the previous example using bitstrings:
4.8.5 Using DNAME for Delegation of IPv6 Reverse Addresses
Delegation of reverse addresses is done through the new DNAME RR. In the
example above, where \[x2/3]
.ip6.int. needs to delegate \[xFFF0] to an
organization (
example2.com ), the domain administrator would insert a line
similar to the following in the \[x2/3]
.ip6.int. zone:
We suggest that the top of your administrative control (
example.com , in
this case) provide all the bits required for reverse and forward resolution
to allow name resolution even if the network is disconnected from the
Internet. This will also allow operation with DNSSEC if you set up a false
trusted server for "." containing only delegations for your forward and
reverse zones directly to the top of your administrative control. This
should be signed with a key trusted by all of your clients, equivalent to
------------------------------------------------------------------------
Section 5. The BIND 9 Lightweight Resolver
5.1 The Lightweight Resolver Library
Traditionally applications have been linked with a stub resolver library
that sends recursive DNS queries to a local caching name server.
IPv6 introduces new complexity into the resolution process, such as
following A6 chains and DNAME records, and simultaneous lookup of IPv4 and
IPv6 addresses. These are hard or impossible to implement in a traditional
Instead, BIND 9 provides resolution services to local clients using a
combination of a lightweight resolver library and a resolver daemon process
running on the local host. These communicate using a simple UDP-based
protocol, the "lightweight resolver protocol" that is distinct from and
simpler than the full DNS protocol.
5.2 Running a Resolver Daemon
To use the lightweight resolver interface, the system must run the resolver
Applications using the lightweight resolver library will make UDP requests
to the IPv4 loopback address (127.0.0.1) on port 921. The daemon will try to
find the answer to the questions "what are the addresses for host
foo.example.com ?" and "what are the names for IPv4 address 204.152.184.79?"
The daemon currently only looks in the DNS, but in the future it may use
The lwresd daemon is essentially a stripped-down, caching-only name server
that answers requests using the lightweight resolver protocol rather than
the DNS protocol. Because it needs to run on each host, it is designed to
require no or minimal configuration. It uses the name servers listed on
doing the resolution autonomously if none are specified.
------------------------------------------------------------------------
Section 6. BIND 9 Configuration Reference
BIND 9 configuration is broadly similar to BIND
8.x; however, there are a
few new areas of configuration, such as views. BIND
8.x configuration files
should work with few alterations in BIND 9, although more complex
configurations should be reviewed to check if they can be more efficiently
implemented using the new features found in BIND 9.
BIND 4 configuration files can be converted to the new format using the
6.1 Configuration File Elements
Following is a list of elements used throughout the BIND configuration file
acl_name The name of an address_match_list as defined by the acl
address_match_list A list of one or more ip_addr , ip_prefix , key_id , or
acl_name elements, as described in Address Match Lists
domain_name A quoted string which will be used as a DNS name, for
dotted_decimal One or more integers valued 0 through 255 separated
only by dots (`.'), such as 123 , 45.67 or 89.123.45.67
ip4_addr An IPv4 address with exactly four elements in
An IPv6 address, such as fe80::200:f8ff:fe01:9742 .
An ip4_addr or ip6_addr .
An IP port number . number is limited to 0 through
ip_port 65535, with values below 1024 typically restricted to
root-owned processes. In some cases an asterisk (`*')
character can be used as a placeholder to select a
random high-numbered port.
An IP network specified as an ip_addr , followed by a
ip_prefix slash (`/') and then the number of bits in the netmask.
For example, 127/8 is the network 127.0.0.0 with
netmask 255.0.0.0 and 1.2.3.0/28 is network 1.2.3.0
with netmask 255.255.255.240 .
key_name A domain_name representing the name of a shared key, to
be used for transaction security.
A non-negative integer with an entire range limited by
number the range of a C language signed integer (2,147,483,647
on a machine with 32 bit integers). Its acceptable
value might further be limited by the context in which
path_name A quoted string which will be used as a pathname, such
A number, the word unlimited , or the word default .
The maximum value of size_spec is that of unsigned long
integers on the machine. An unlimited size_spec
requests unlimited use, or the maximum available
amount. A default size_spec uses the limit that was in
force when the server was started.
size_spec A number can optionally be followed by a scaling
factor: K or k for kilobytes, M or m for megabytes, and
G or g for gigabytes, which scale by 1024, 1024*1024,
and 1024*1024*1024 respectively.
Integer storage overflow is currently silently ignored
during conversion of scaled values, resulting in values
less than intended, possibly even negative. Using
unlimited is the best way to safely set a really large
yes_or_no Either yes or no . The words true and false are also
accepted, as are the numbers 1 and 0 .
6.1.1 Address Match Lists
address_match_list = address_match_list_element ;
[ address_match_list_element; ... ]
address_match_list_element = [ ! ] (ip_address [/length] |
key key_id | acl_name | { address_match_list } )
6.1.1.2 Definition and Usage
Address match lists are primarily used to determine access control for
various server operations. They are also used to define priorities for
querying other nameservers and to set the addresses on which named will
listen for queries. The elements which constitute an address match list can
* an IP address (IPv4 or IPv6)
* an IP prefix (in the `/'-notation)
* a key ID, as defined by the key statement
* the name of an address match list previously defined with the acl
* a nested address match list enclosed in braces
Elements can be negated with a leading exclamation mark (`!') and the match
list names "any," "none," "localhost" and "localnets" are predefined. More
information on those names can be found in the description of the acl
The addition of the key clause made the name of this syntactic element
something of a misnomer, since security keys can be used to validate access
without regard to a host or network address. Nonetheless, the term "address
match list" is still used throughout the documentation.
When a given IP address or prefix is compared to an address match list, the
list is traversed in order until an element matches. The interpretation of a
match depends on whether the list is being used for access control, defining
listen-on ports, or as a topology, and whether the element was negated.
When used as an access control list, a non-negated match allows access and a
negated match denies access. If there is no match, access is denied. The
clauses allow-query , allow-transfer , allow-update and blackhole all use
address match lists this. Similarly, the listen-on option will cause the
server to not accept queries on any of the machine's addresses which do not
When used with the topology clause, a non-negated match returns a distance
based on its position on the list (the closer the match is to the start of
the list, the shorter the distance is between it and the server). A negated
match will be assigned the maximum distance from the server. If there is no
match, the address will get a distance which is further than any non-negated
list element, and closer than any negated element.
Because of the first-match aspect of the algorithm, an element that defines
a subset of another element in the list should come before the broader
element, regardless of whether either is negated. For example, in
1.2.3/24; ! 1.2.3.13; the 1.2.3.13 element is completely useless because the
algorithm will match any lookup for 1.2.3.13 to the 1.2.3/24 element. Using
! 1.2.3.13; 1.2.3/24 fixes that problem by having 1.2.3.13 blocked by the
negation but all other 1.2.3.* hosts fall through.
The BIND 9 comment syntax allows for comments to appear anywhere that white
space may appear in a BIND configuration file. To appeal to programmers of
all kinds, they can be written in C, C++, or
shell/perl constructs.
/* This is a BIND comment as in C */
// This is a BIND comment as in C++
# This is a BIND comment as in common UNIX shells and perl
6.1.2.2 Definition and Usage
Comments may appear anywhere that whitespace may appear in a BIND
C-style comments start with the two characters /* (slash, star) and end with
*/ (star, slash). Because they are completely delimited with these
characters, they can be used to comment only a portion of a line or to span
C-style comments cannot be nested. For example, the following is not valid
because the entire comment ends with the first */:
/* This is the start of a comment.
This is still part of the comment.
/* This is an incorrect attempt at nesting a comment. */
This is no longer in any comment. */
C++-style comments start with the two characters // (slash, slash) and
continue to the end of the physical line. They cannot be continued across
multiple physical lines; to have one logical comment span multiple lines,
each line must use the // pair.
// This is the start of a comment. The next line
// is a new comment, even though it is logically
// part of the previous comment.
Shell-style (or perl-style, if you prefer) comments start with the character
# (number sign) and continue to the end of the physical line, as in C++
# This is the start of a comment. The next line
# is a new comment, even though it is logically
# part of the previous comment.
WARNING: you cannot use the semicolon (`;') character to start a comment
such as you would in a zone file. The semicolon indicates the end of a
6.2 Configuration File Grammar
A BIND 9 configuration consists of statements and comments. Statements end
with a semicolon. Statements and comments are the only elements that can
appear without enclosing braces. Many statements contain a block of
substatements, which are also terminated with a semicolon.
The following statements are supported:
acl defines a named IP address matching list, for access control
controls declares control channels to be used by the rndc utility.
key specifies key information for use in authentication and
authorization using TSIG.
logging specifies what the server logs, and where the log messages
options controls global server configuration options and sets
defaults for other statements.
server sets certain configuration options on a per-server basis.
trusted-keys defines trusted DNSSEC keys.
The logging and options statements may only occur once per configuration.
6.2.1 acl Statement Grammar
6.2.2 acl Statement Definition and Usage
The acl statement assigns a symbolic name to an address match list. It gets
its name from a primary use of address match lists: Access Control Lists
Note that an address match list's name must be defined with acl before it
can be used elsewhere; no forward references are allowed.
The following ACLs are built-in:
localhost Matches the IP addresses of all interfaces on the system.
localnets Matches any host on a network for which the system has an
6.2.3 controls Statement Grammar
[ inet (ip_addr|*) port ip_port allow { address_match_list } ;
[ unix string permission number owner number group number ;
6.2.4 controls Statement Definition and Usage
The controls statement declares control channels to be used by system
administrators to affect the operation of the local nameserver. These
control channels are used by the ndc utility to send commands to and
retrieve non-DNS results from a nameserver.
A UNIX control channel is a "first in first out" (FIFO) named pipe in the
file system, and access to it is controlled by normal file system
permissions. It is created by named with the specified file mode bits (see
the chmod(1) manual page), user and group owner. Note that, unlike chmod ,
the mode bits specified for permission will normally have a leading 0 so the
number is interpreted as octal. Also note that the user and group ownership
specified as owner and group must be given as numbers, not names. It is
recommended that the permissions be restricted to administrative personnel
only to prevent random users on the system from having the ability to manage
An inet control channel is a
TCP/IP socket accessible to the Internet,
created at the specified ip_port on the specified ip_addr . It is
recommended that 127.0.0.1 be the only ip_addr used, and this only if you
trust all non-privileged users on the local host to manage your nameserver.
The controls statement is not yet implemented in BIND 9. The server always
listens for control connections on IP address 127.0.0.1, port 953.
6.2.5 include Statement Grammar
6.2.6 include Statement Definition and Usage
The include statement inserts the specified file at the point that the
include statement is encountered. The include statement facilitates the
administration of configuration files by permitting the reading or writing
of some things but not others. For example, the statement could include
private keys that are readable only by a nameserver.
6.2.7 key Statement Grammar
6.2.8 key Statement Definition and Usage
The key statement defines a shared secret key for use with TSIG. See TSIG .
The key_id , also known as the key name, is a domain name uniquely
identifying the key. It can be used in a "server" statement to cause
requests sent to that server to be signed with this key, or in address match
lists to verify that incoming requests have been signed with a key matching
this name, algorithm, and secret.
algorithm. The only algorithm currently supported with TSIG authentication
is hmac-md5 . The secret_string is the secret to be used by the algorithm,
and is treated as a base-64 encoded string.
6.2.9 logging Statement Grammar
[ versions ( number | unlimited ) ]
| syslog ( syslog_facility
[ severity (critical | error | warning | notice |
info | debug [ level ] | dynamic ; ]
[ print-category yes or no;
[ print-severity yes or no; ]
[ print-time yes or no; ]
[ category category_name {
channel_name ; [ channel_name ; ... ]
6.2.10 logging Statement Definition and Usage
The logging statement configures a wide variety of logging options for the
nameserver. Its channel phrase associates output methods, format options and
severity levels with a name that can then be used with the category phrase
to select how various classes of messages are logged.
Only one logging statement is used to define as many channels and categories
as are wanted. If there is no logging statement, the logging configuration
category "default" { "default_syslog"; "default_debug"; };
In BIND 9, the logging configuration is only established when the entire
configuration file has been parsed. In BIND 8, it was established as soon as
the logging statement was parsed. When the server is starting up, all
logging messages regarding syntax errors in the configuration file go to the
default channels, or to standard error if the " -g " option was specified.
6.2.10.1 The channel Phrase
All log output goes to one or more channels ; you can make as many of them
Every channel definition must include a clause that says whether messages
selected for the channel go to a file, to a particular syslog facility, or
are discarded. It can optionally also limit the message severity level that
will be accepted by the channel (the default is info ), and whether to
include a named -generated time stamp, the category name
and/or severity
level (the default is not to include any).
The word null as the destination option for the channel will cause all
messages sent to it to be discarded; in that case, other options for the
The file clause can include limitations both on how large the file is
allowed to become, and how many versions of the file will be saved each time
The size option for files is simply a hard ceiling on log growth. If the
file ever exceeds the size, then named will not write anything more to it
until the file is reopened; exceeding the size does not automatically
trigger a reopen. The default behavior is not to limit the size of the file.
If you use the version log file option, then named will retain that many
backup versions of the file by renaming them when opening. For example, if
you choose to keep 3 old versions of the file
lamers.log then just before it
versions are kept by default; any existing log file is simply appended. The
unlimited keyword is synonymous with 99 in current BIND releases.
Example usage of the size and versions options:
channel "an_example_channel" {
The argument for the syslog clause is a syslog facility as described in the
syslog man page. How syslog will handle messages sent to this facility is
described in the
syslog.conf man page. If you have a system which uses a
very old version of syslog that only uses two arguments to the openlog()
function, then this clause is silently ignored.
The severity clause works like syslog 's "priorities," except that they can
also be used if you are writing straight to a file rather than using syslog
. Messages which are not at least of the severity level given will not be
selected for the channel; messages of higher severity levels will be
If you are using syslog , then the
syslog.conf priorities will also
determine what eventually passes through. For example, defining a channel
facility and severity as daemon and debug but only logging
daemon.warningvia
syslog.conf will cause messages of severity info and notice to be
dropped. If the situation were reversed, with named writing messages of only
warning or higher, then syslogd would print all messages it received from
The server can supply extensive debugging information when it is in
debugging mode. If the server's global debug level is greater than zero,
then debugging mode will be active. The global debug level is set either by
starting the named server with the " -d " flag followed by a positive
integer, or by running rndc trace ( the latter method is not yet implemented
). The global debug level can be set to zero, and debugging mode turned off,
by running ndc notrace . All debugging messages in the server have a debug
level, and higher debug levels give more detailed output. Channels that
specify a specific debug severity, for example:
channel "specific_debug_level" {
will get debugging output of level 3 or less any time the server is in
debugging mode, regardless of the global debugging level. Channels with
dynamic severity use the server's global level to determine what messages to
If print-time has been turned on, then the date and time will be logged.
print-time may be specified for a syslog channel, but is usually pointless
since syslog also prints the date and time. If print-category is requested,
then the category of the message will be logged as well. Finally, if
print-severity is on, then the severity level of the message will be logged.
The print- options may be used in any combination, and will always be
printed in the following order: time, category, severity. Here is an example
where all three print- options are on:
28-Feb-2000 15:05:32.863 general: notice: running
There are four predefined channels that are used for named 's default
logging as follows. How they are used is described in the category Phrase .
channel "default_syslog" {
syslog daemon; // end to syslog's daemon
severity info; // only send priority info
channel "default_debug" {
// Note: stderr is used instead
// if the server is started
severity dynamic // log at the server's
channel "default_stderr" { // writes to stderr
file "<stderr>"; // this is illustrative only;
// there's currently no way of
// specifying an internal file
// configuration language.
severity info; // only send priority info
null; // toss anything sent to
The default_debug channel normally writes to a file
named.run in the
server's working directory. For security reasons, when the " -u " command
line option is used, the
named.run file is created only after named has
changed to the new UID, and any debug output generated while named is
starting up and still running as root is discarded. If you need to capture
this output, you must run the server with the " -g " option and redirect
standard error to a file.
Once a channel is defined, it cannot be redefined. Thus you cannot alter the
built-in channels directly, but you can modify the default logging by
pointing categories at channels you have defined.
6.2.10.2 The category Phrase
There are many categories, so you can send the logs you want to see wherever
you want, without seeing logs you don't want. If you don't specify a list of
channels for a category, then log messages in that category will be sent to
the default category instead. If you don't specify a default category, the
following "default default" is used:
category "default" { "default_syslog"; "default_debug"; };
As an example, let's say you want to log security events to a file, but you
also want keep the default logging behavior. You'd specify the following:
channel "my_security_channel" {
To discard all messages in a category, specify the null channel:
category "xfer-out" { "null"; };
category "notify" { "null"; };
Following are the available categories and brief descriptions of the types
of log information they contain . More categories may be added in future
default The default category defines the logging options for those
categories where no specific configuration has been defined.
general The catch-all. Many things still aren't classified into
categories, and they all end up here.
database Messages relating to the databases used internally by the name
server to store zone and cache data.
security Approval and denial of requests.
config Configuration file parsing and processing.
resolver DNS resolution, such as the recursive lookups performed on behalf
of clients by a caching name server.
xfer-in Zone transfers the server is receiving.
xfer-out Zone transfers the server is sending.
notify The NOTIFY protocol.
client Processing of client requests.
network Network operations.
6.2.11 options Statement Grammar
This is the grammar of the option statement in the
named.conf file:
[ version version_string; ]
[ named-xfer path_name; ]
[ tkey-domain domainname; ]
[ tkey-dhkey keyname keyid; ]
[ memstatistics-file path_name; ]
[ statistics-file path_name; ]
[ auth-nxdomain yes_or_no; ]
[ deallocate-on-exit yes_or_no; ]
[ fake-iquery yes_or_no; ]
[ fetch-glue yes_or_no; ]
[ has-old-clients yes_or_no; ]
[ host-statistics yes_or_no; ]
[ multiple-cnames yes_or_no; ]
[ rfc2308-type1 yes_or_no; ]
[ use-id-pool yes_or_no; ]
[ maintain-ixfr-base yes_or_no; ]
[ forward ( only | first ); ]
[ forwarders { [ in_addr ; [ in_addr ; ... ] ] }; ]
[ check-names ( master | slave | response )( warn | fail | ignore ); ]
[ allow-query { address_match_list }; ]
[ allow-transfer { address_match_list }; ]
[ allow-recursion { address_match_list }; ]
[ blackhole { address_match_list }; ]
[ listen-on [ port ip_port ] { address_match_list }; ]
[ query-source [ address ( ip_addr | * ) ] [ port ( ip_port | * ) ]; ]
[ max-transfer-time-in number; ]
[ max-transfer-time-out number; ]
[ max-transfer-idle-in number; ]
[ max-transfer-idle-out number; ]
[ recursive-clients number; ]
[ serial-queries number; ]
[ transfer-format ( one-answer | many-answers ); ]
[ transfers-out number; ]
[ transfers-per-ns number; ]
[ transfer-source ip4_addr; ]
[ transfer-source-v6 ip6_addr; ]
[ also-notify { ip_addr; [ ip_addr; ... ] }; ]
[ max-ixfr-log-size number; ] [ coresize size_spec ; ] [ datasize size_spec ; ] [ files size_spec ; ] [ stacksize size_spec ; ] [ cleaning-interval number; ] [ heartbeat-interval number; ] [ interface-interval number; ] [ statistics-interval number; ]
[ topology { address_match_list }; ]
[ sortlist { address_match_list }; ]
[ rrset-order { order_spec ; [ order_spec ; ... ] ] }; [ lame-ttl number; ] [ max-ncache-ttl number; ]
[ max-cache-ttl number; ]
[ sig-validity-interval number ; ]
[ treat-cr-as-space yes_or_no ; ]
6.2.12 options Statement Definition and Usage
The options statement sets up global options to be used by BIND. This
statement may appear only once in a configuration file. If more than one
occurrence is found, the first occurrence determines the actual options
used, and a warning will be generated. If there is no options statement, an
options block with each option set to its default will be used.
The version the server should report via a query of
version name
version.bind in class chaos . The default is the
real version number of this server.
The working directory of the server. Any non-absolute
pathnames in the configuration file will be taken as
relative to this directory. The default location for
directory. If a directory is not specified, the working
directory defaults to ` . ', the directory from which
the server was started. The directory specified should
This option is obsolete. It was used in BIND 8 to
named-xfer specify the pathname to the named-xfer program. In
BIND 9, no separate named-xfer program is needed; its
functionality is built into the name server.
The domain appended to the names of all shared keys
generated with TKEY . When a client requests a TKEY
exchange, it may or may not specify the desired name
tkey-domain for the key. If present, the name of the shared key
will be " client specified part " + " tkey-domain ".
Otherwise, the name of the shared key will be " random
hex digits " + " tkey-domain ". In most cases, the
domainname should be the server's domain name.
The Diffie-Hellman key used by the server to generate
shared keys with clients using the Diffie-Hellman mode
tkey-dhkey of TKEY . The server must be able to load the public
and private keys from files in the working directory.
In most cases, the keyname should be the server's host
The pathname of the file the server dumps the database
dump-file to when it receives SIGINT signal ( ndc dumpdb ). If
The pathname of the file the server writes memory usage
memstatistics-file statistics to on exit. If not specified, the default is
The pathname of the file the server writes its process
ID in. If not specified, the default is operating
pid-file system dependent, but is usually
used by programs that want to send signals to the
The pathname of the file the server appends statistics
statistics-file to. If not specified, the default is
named.stats . Not
yet implemented in BIND 9 .
If yes , then the AA bit is always set on NXDOMAIN
responses, even if the server is not actually
auth-nxdomain authoritative. The default is no ; this is a change
from BIND 8. If you are using very old DNS software,
you may need to set it to yes .
This option was used in BIND 8 to enable checking for
deallocate-on-exit memory leaks on exit. BIND 9 ignores the option and
always performs the checks.
If yes , then the server treats all zones as if they
are doing zone transfers across a dial on demand dialup
link, which can be brought up by traffic originating
from this server. This has different effects according
to zone type and concentrates the zone maintenance so
that it all happens in a short interval, once every
heartbeat-interval and hopefully during the one call.
It also suppresses some of the normal zone maintenance
traffic. The default is no .
The dialup option may also be specified in the zone
statement, in which case it overrides the options
If the zone is a master then the server will send out a
NOTIFY request to all the slaves. This will trigger the
zone serial number check in the slave (providing it
supports NOTIFY) allowing the slave to verify the zone
while the connection is active.
If the zone is a slave or stub then the server will
suppress the regular "zone up to date" queries and only
heartbeat-interval expires. Not yet implemented in
In BIND 8, this option was used to enable simulating
fake-iquery the obsolete DNS query type IQUERY. BIND 9 never does
(Information present outside of the authoritative nodes
in the zone is called glue information). If yes (the
default), the server will fetch glue resource records
it doesn't have when constructing the additional data
fetch-glue section of a response. fetch-glue no can be used in
conjunction with recursion no to prevent the server's
cache from growing or becoming corrupted (at the cost
of requiring more work from the client). Not yet
This option was incorrectly implemented in BIND 8, and
has-old-clients is ignored by BIND 9. To achieve the intended effect of
has-old-clients yes , specify the two separate options
auth-nxdomain yes and rfc2308-type1 no instead.
If yes , then statistics are kept for every host that
host-statistics the nameserver interacts with. The default is no .
Note: turning on host-statistics can consume huge
amounts of memory. Not yet implemented in BIND 9.
This option is obsolete . It was used in BIND 8 to
determine whether a transaction log was kept for
maintain-ixfr-base Incremental Zone Transfer. BIND 9 maintains a
transaction log whenever possible. If you need to
disable outgoing incremental zone transfers, use
This option was used in BIND 8 to allow a domain name
to allow multiple CNAME records in violation of the DNS
standards. BIND 9 currently does not check for multiple
multiple-cnames CNAMEs in zone data loaded from master files, but such
checks may be introduced in a later release. BIND 9
always strictly enforces the CNAME rules in dynamic
If yes (the default), DNS NOTIFY messages are sent when
a zone the server is authoritative for changes. See
Notify , for more information. The notify option may
notify also be specified in the zone statement, in which case
it overrides the options notify statement. It would
only be necessary to turn off this option if it caused
If yes , and a DNS query requests recursion, then the
server will attempt to do all the work required to
recursion answer the query. If recursion is not on, the server
will return a referral to the client if it doesn't know
the answer. The default is yes . See also fetch-glue
Setting this to yes will cause the server to send NS
rfc2308-type1 records along with the SOA record for negative answers.
The default is no . Not yet implemented in BIND 9 .
use-id-pool This option is obsolete . BIND 9 always allocates query
This option was used in BIND 8 to make the server treat
" \r " characters the same way as <space> " " or " \t
treat-cr-as-space ", to facilitate loading of zone files on a UNIX system
that were generated on an NT or DOS machine. In BIND 9,
both UNIX " \n " and
NT/DOS " \r\n " newlines are
always accepted, and the option is ignored.
The forwarding facility can be used to create a large site-wide cache on a
few servers, reducing traffic over links to external nameservers. It can
also be used to allow queries by servers that do not have direct access to
the Internet, but wish to look up exterior names anyway. Forwarding occurs
only on those queries for which the server is not authoritative and does not
have the answer in its cache.
This option is only meaningful if the forwarders list is not
empty. A value of first , the default, causes the server to
forward query the forwarders first, and if that doesn't answer the
question the server will then look for the answer itself. If
only is specified, the server will only query the forwarders.
forwarders Specifies the IP addresses to be used for forwarding. The
default is the empty list (no forwarding).
Forwarding can also be configured on a per-domain basis, allowing for the
global forwarding options to be overridden in a variety of ways. You can set
particular domains to use different forwarders, or have a different forward
only/first behavior, or not forward at all. See zone Statement Grammar for
The server can check domain names based upon their expected client contexts.
For example, a domain name used as a hostname can be checked for compliance
with the RFCs defining valid hostnames.
Three checking methods are available:
ignore No checking is done.
warn Names are checked against their expected client contexts. Invalid
names are logged, but processing continues normally.
fail Names are checked against their expected client contexts. Invalid
names are logged, and the offending data is rejected.
The server can check names in three areas: master zone files, slave zone
files, and in responses to queries the server has initiated. If check-names
response fail has been specified, and answering the client's question would
require sending an invalid name to the client, the server will send a
REFUSED response code to the client.
check-names response ignore;
check-names may also be specified in the zone statement, in which case it
overrides the options check-names statement. When used in a zone statement,
the area is not specified because it can be deduced from the zone type.
Name checking is not yet implemented in BIND 9.
Access to the server can be restricted based on the IP address of the
requesting system. See Address Match Lists for details on how to specify IP
Specifies which hosts are allowed to ask ordinary
questions. allow-query may also be specified in the zone
allow-query statement, in which case it overrides the options
allow-query statement. If not specified, the default is to
allow queries from all hosts.
Specifies which hosts are allowed to make recursive
allow-recursion queries through this server. If not specified, the default
is to allow recursive queries from all hosts.
Specifies which hosts are allowed to receive zone
transfers from the server. allow-transfer may also be
allow-transfer specified in the zone statement, in which case it
overrides the options allow-transfer statement. If not
specified, the default is to allow transfers from all
Specifies a list of addresses that the server will not
blackhole accept queries from or use to resolve a query. Queries
from these addresses will not be responded to. The default
is none . Not yet implemented in BIND 9.
The interfaces and ports that the server will answer queries from may be
specified using the listen-on option. listen-on takes an optional port, and
an address_match_list . The server will listen on all interfaces allowed by
the address match list. If a port is not specified, port 53 will be used.
Multiple listen-on statements are allowed. For example,
listen-on port 1234 { !1.2.3.4; 1.2/16; };
will enable the nameserver on port 53 for the IP address 5.6.7.8, and on
port 1234 of an address on the machine in net 1.2 that is not 1.2.3.4.
If no listen-on is specified, the server will listen on port 53 on all
The listen-on-v6 option is used to specify the ports on which the server
will listen for incoming queries sent using IPv6.
The server does not bind a separate socket to each IPv6 interface address as
it does for IPv4. Instead, it always listens on the IPv6 wildcard address.
Therefore, the only values allowed for the address_match_list argument to
the listen-on-v6 statement are " { any; } " and " { none; } ".
Multiple listen-on-v6 options can be used to listen on multiple ports:
listen-on-v6 port 53 { any; };
listen-on-v6 port 1234 { any; };
To make the server not listen on any IPv6 address, use
If no listen-on-v6 statement is specified, the server will listen on port 53
on the IPv6 wildcard address.
If the server doesn't know the answer to a question, it will query other
nameservers. query-source specifies the address and port used for such
queries. For queries sent over IPv6, there is a separate query-source-v6
option. If address is * or is omitted, a wildcard IP address ( INADDR_ANY )
will be used. If port is * or is omitted, a random unprivileged port will be
query-source address * port *;
query-source-v6 address * port *
Note: query-source currently applies only to UDP queries; TCP queries always
use a wildcard IP address and a random unprivileged port.
BIND has mechanisms in place to facilitate zone transfers and set limits on
the amount of load that transfers place on the system. The following options
Defines a global list of IP addresses that are also
sent NOTIFY messages whenever a fresh copy of the
zone is loaded. This helps to ensure that copies of
the zones will quickly converge on stealth servers.
also-notify If an also-notify list is given in a zone statement,
it will override the options also-notify statement.
When a zone notify statement is set to no , the IP
addresses in the global also-notify list will not be
sent NOTIFY messages for that zone. The default is
the empty list (no global notification list).
Inbound zone transfers running longer than this many
max-transfer-time-in minutes will be terminated. The default is 120
Inbound zone transfers making no progress in this
max-transfer-idle-in many minutes will be terminated. The default is 60
Outbound zone transfers running longer than this
max-transfer-time-out many minutes will be terminated. The default is 120
Outbound zone transfers making no progress in this
many minutes will be terminated. The default is 60
max-transfer-idle-out minutes
Slave servers will periodically query master servers
to find out if zone serial numbers have changed.
Each such query uses a minute amount of the slave
server's network bandwidth, but more importantly
each query uses a small amount of memory in the
slave server while waiting for the master server to
respond. The serial-queries option sets the maximum
number of concurrent serial-number queries allowed
to be outstanding at any given time. The default is
serial-queries 4. Note: If a server loads a large (tens or hundreds
of thousands) number of slave zones, then this limit
should be raised to the high hundreds or low
thousands, otherwise the slave server may never
actually become aware of zone changes in the master
servers. Beware, though, that setting this limit
arbitrarily high can spend a considerable amount of
your slave server's network, CPU, and memory
resources. As with all tunable limits, this one
should be changed gently and monitored for its
effects. Not yet implemented in BIND 9.
The server supports two zone transfer methods.
one-answer uses one DNS message per resource record
transferred. many-answers packs as many resource
records as possible into a message. many-answers is
transfer-format more efficient, but is only known to be understood
by BIND 9, BIND
8.x and patched versions of
BIND 4.9.5. The default is many-answers .
transfer-format may be overridden on a per-server
basis by using the server statement.
The maximum number of inbound zone transfers that
can be running concurrently. The default value is 10
transfers-in . Increasing transfers-in may speed up the
convergence of slave zones, but it also may increase
the load on the local system.
The maximum number of outbound zone transfers that
transfers-out can be running concurrently. Zone transfer requests
in excess of the limit will be refused. The default
The maximum number of inbound zone transfers that
can be concurrently transferring from a given remote
nameserver. The default value is 2 . Increasing
transfers-per-ns transfers-per-ns may speed up the convergence of
slave zones, but it also may increase the load on
the remote nameserver. transfers-per-ns may be
overridden on a per-server basis by using the
transfers phrase of the server statement.
transfer-source determines which local address will
be bound to IPv4 TCP connections used to fetch zones
transferred inbound by the server. If not set, it
defaults to a system controlled value which will
usually be the address of the interface "closest to"
the remote end. This address must appear in the
transfer-source remote end's allow-transfer option for the zone
being transferred, if one is specified. This
statement sets the transfer-source for all zones,
but can be overridden on a per-zone basis by
transfer-source statement within the zone block in
transfer-source-v6 The same as transfer-source , except zone transfers
are performed using IPv6.
The server's usage of many system resources can be limited. Some operating
systems don't support some of the limits. On such systems, a warning will be
issued if the unsupported limit is used. Some operating systems don't
support limiting resources.
Scaled values are allowed when specifying resource limits. For example, 1G
can be used instead of 1073741824 to specify a limit of one gigabyte.
unlimited requests unlimited use, or the maximum available amount. default
uses the limit that was in force when the server was started. See the
description of size_spec in Configuration File Elements for more details.
coresize The maximum size of a core dump. The default is default
. Not yet implemented in BIND 9.
datasize The maximum amount of data memory the server may use.
The default is default . Not yet implemented in BIND 9.
The maximum number of files the server may have open
concurrently. The default is unlimited . Note: on some
operating systems the server cannot set an unlimited
value and cannot determine the maximum number of open
files files the kernel can support. On such systems, choosing
unlimited will cause the server to use the larger of the
rlim_max for RLIMIT_NOFILE and the value returned by
sysconf(_SC_OPEN_MAX) . If the actual kernel limit is
larger than this value, use limit files to specify the
limit explicitly. Not yet implemented in BIND 9.
The max-ixfr-log-size will be used in a future release
max-ixfr-log-size of the server to limit the size of the transaction log
kept for Incremental Zone Transfer. Not yet implemented
The maximum number of simultaneous recursive lookups the
recursive-clients server will perform on behalf of clients. The default is
stacksize The maximum amount of stack memory the server may use.
The default is default . Not yet implemented in BIND 9.
The maximum number of simultaneous client TCP
tcp-clients connections that the server will accept. The default is
Resource limits are not yet implemented in BIND 9.
6.2.12.9 Periodic Task Intervals
The server will remove expired resource records from
cleaning-interval the cache every cleaning-interval minutes. The default
is 60 minutes. If set to 0 , no periodic cleaning will
The server will perform zone maintenance tasks for all
zones marked dialup yes whenever this interval
heartbeat-interval expires. The default is 60 minutes. Reasonable values
are up to 1 day (1440 minutes). If set to 0 , no zone
maintenance for these zones will occur. Not yet
The server will scan the network interface list every
interface-interval minutes. The default is 60 minutes.
If set to 0 , interface scanning will only occur when
interface-interval the configuration file is loaded. After the scan,
listeners will be started on any new interfaces
(provided they are allowed by the listen-on
configuration). Listeners on interfaces that have gone
Nameserver statistics will be logged every
statistics-interval statistics-interval minutes. The default is 60 . If
set to 0 , no statistics will be logged. Not yet
All other things being equal, when the server chooses a nameserver to query
from a list of nameservers, it prefers the one that is topologically closest
to itself. The topology statement takes an address_match_list and interprets
it in a special way. Each top-level list element is assigned a distance.
Non-negated elements get a distance based on their position in the list,
where the closer the match is to the start of the list, the shorter the
distance is between it and the server. A negated match will be assigned the
maximum distance from the server. If there is no match, the address will get
a distance which is further than any non-negated list element, and closer
than any negated element. For example,
will prefer servers on network 10 the most, followed by hosts on network
1.2.0.0 (netmask 255.255.0.0) and network 3, with the exception of hosts on
network 1.2.3 (netmask 255.255.255.0), which is preferred least of all.
topology { localhost; localnets; };
The topology option is not yet implemented in BIND 9.
6.2.12.11 The sortlist Statement
Resource Records (RRs) are the data associated with the names in a domain
name space. The data is maintained in the form of sets of RRs. The order of
RRs in a set is, by default, not significant. Therefore, to control the
sorting of records in a set of resource records, or RRset , you must use the
RRs are explained more fully in See Types of Resource Records and When to
Use Them . Specifications for RRs are documented in RFC 1035.
When returning multiple RRs the nameserver will normally return them in
Round Robin order, that is, after each request the first RR is put at the
end of the list. The client resolver code should rearrange the RRs as
appropriate, that is, using any addresses on the local net in preference to
other addresses. However, not all resolvers can do this or are correctly
configured. When a client is using a local server the sorting can be
performed in the server, based on the client's address. This only requires
configuring the nameservers, not all the clients.
The sortlist statement (see below) takes an address_match_list and
interprets it even more specifically than the topology statement does (see
Topology ). Each top level statement in the sortlist must itself be an
explicit address_match_list with one or two elements. The first element
(which may be an IP address, an IP prefix, an ACL name or a nested
address_match_list ) of each top level list is checked against the source
address of the query until a match is found.
Once the source address of the query has been matched, if the top level
statement contains only one element, the actual primitive element that
matched the source address is used to select the address in the response to
move to the beginning of the response. If the statement is a list of two
elements, then the second element is treated the same as the
address_match_list in a topology statement. Each top level element is
assigned a distance and the address in the response with the minimum
distance is moved to the beginning of the response.
In the following example, any queries received from any of the addresses of
the host itself will get responses preferring addresses on any of the
locally connected networks. Next most preferred are addresses on the
192.168.1/24 network, and after that either the 192.168.2/24 or
192.168.3/24 network with no preference shown between these two networks.
Queries received from a host on the 192.168.1/24 network will prefer other
addresses on that network to the 192.168.2/24 and
192.168.3/24 networks. Queries received from a host on the 192.168.4/24 or
the 192.168.5/24 network will only prefer other addresses on their directly
{ localhost; // IF the local host
{ localnets; // THEN first fit on the
192.168.1/24; // following nets
{ 192,168.2/24; 192.168.3/24; }; }; };
{ 192.168.1/24; // IF on class C 192.168.1
{ 192.168.1/24; // THEN use .1, or .2 or .3
{ 192.168.2/24; 192.168.3/24; }; }; };
{ 192.168.2/24; // IF on class C 192.168.2
{ 192.168.2/24; // THEN use .2, or .1 or .3
{ 192.168.1/24; 192.168.3/24; }; }; };
{ 192.168.3/24; // IF on class C 192.168.3
{ 192.168.3/24; // THEN use .3, or .1 or .2
{ 192.168.1/24; 192.168.2/24; }; }; };
{ { 192.168.4/24; 192.168.5/24; };
// if .4 or .5, prefer that net
The following example will give reasonable behavior for the local host and
hosts on directly connected networks. It is similar to the behavior of the
address sort in BIND
8.x. Responses sent to queries from the local host will
favor any of the directly connected networks. Responses sent to queries from
any other hosts on a directly connected network will prefer addresses on
that same network. Responses to other queries will not be sorted.
{ localhost; localnets; };
The sortlist option is not yet implemented in BIND 9.
When multiple records are returned in an answer it may be useful to
configure the order of the records placed into the response. For example,
the records for a zone might be configured always to be returned in the
order they are defined in the zone file. Or perhaps a random shuffle of the
records as they are returned is wanted. The rrset-order statement permits
configuration of the ordering made of the records in a multiple record
response. The default, if no ordering is defined, is a cyclic ordering
An order_spec is defined as follows:
[ class class_name ][ type type_name ][ name "domain_name"]
If no class is specified, the default is ANY . If no type is specified, the
default is ANY . If no name is specified, the default is " * ".
The legal values for ordering are:
fixed Records are returned in the order they are defined in the zone
random Records are returned in some random order.
cyclic Records are returned in a round-robin order.
will cause any responses for type A records in class IN that have "
other records are returned in cyclic order.
If multiple rrset-order statements appear, they are not combined--the last
If no rrset-order statement is specified, then a default one of:
rrset-order { class ANY type ANY name "*"; order cyclic ;
The rrset-order statement is not yet implemented in BIND 9.
Sets the number of seconds to cache a lame server
indication. 0 disables caching. (This is NOT
lame-ttl recommended.) Default is 600 (10 minutes). Maximum
value is 1800 (30 minutes). Not yet implemented in
To reduce network traffic and increase performance
the server stores negative answers. max-ncache-ttl
is used to set a maximum retention time for these
max-ncache-ttl answers in the server in seconds. The default
max-ncache-ttl is 10800 seconds (3 hours).
max-ncache-ttl cannot exceed 7 days and will be
silently truncated to 7 days if set to a greater
max-cache-ttl sets the maximum time for which the
max-cache-ttl server will cache ordinary (positive) answers. The
default is one week (7 days).
The minimum number of root servers that is required
min-roots for a request for the root servers to be accepted.
Default is 2 . Not yet implemented in BIND 9.
Specifies the number of days into the future when
DNSSEC signatures automatically generated as a
result of dynamic updates (see Dynamic Update ) will
sig-validity-interval expire. The default is 30 days. The signature
inception time is unconditionally set to one hour
before the current time to allow for a limited
6.2.12.14 Deprecated Features
use-ixfr is deprecated in BIND 9. If you need to disable IXFR to a
particular server or servers see the information on the provide-ixfr option
in server Statement Definition and Usage . See also the description of IXFR
in the section Incremental Zone Transfers (IXFR) .
6.2.13 server Statement Grammar
[ provide-ixfr yes_or_no ; ]
[ request-ixfr yes_or_no ; ]
[ transfer-format ( one-answer | many-answers ) ; ]
[ keys { string ; [ string ; [...]] } ; ]
6.2.14 server Statement Definition and Usage
The server statement defines the characteristics to be associated with a
If you discover that a remote server is giving out bad data, marking it as
bogus will prevent further queries to it. The default value of bogus is no .
The bogus clause is not yet implemented in BIND 9.
The provide-ixfr clause determines whether the local server, acting as
master, will respond with an incremental zone transfer when the given remote
server, a slave, requests it. If set to yes , incremental transfer will be
provided whenever possible. If set to no , all transfers to the remote
server will be nonincremental. If not set, the value of the provide-ixfr
option in the global options block is used as a default.
The request-ixfr clause determines whether the local server, acting as a
slave, will request incremental zone transfers from the given remote server,
a master. If not set, the value of the request-ixfr option in the global
options block is used as a default.
IXFR requests to servers that do not support IXFR will automatically fall
back to AXFR. Therefore, there is no need to manually list which servers
support IXFR and which ones do not; the global default of yes should always
work. The purpose of the provide-ixfr and request-ixfr clauses is to make it
possible to disable the use of IXFR even when both master and slave claim to
support it, for example if one of the servers is buggy and crashes or
corrupts data when IXFR is used.
The server supports two zone transfer methods. The first, one-answer , uses
one DNS message per resource record transferred. many-answers packs as many
resource records as possible into a message. many-answers is more efficient,
but is only known to be understood by BIND 9, BIND
8.x, and patched versions
of BIND 4.9.5. You can specify which method to use for a server with the
transfer-format option. If transfer-format is not specified, the
transfer-format specified by the options statement will be used.
transfers is used to limit the number of concurrent inbound zone transfers
from the specified server. If no transfers clause is specified, the limit is
set according to the transfers-per-ns option.
The keys clause is used to identify a key_id defined by the key statement,
to be used for transaction security when talking to the remote server. The
key statement must come before the server statement that references it. When
a request is sent to the remote server, a request signature will be
generated using the key specified here and appended to the message. A
request originating from the remote server is not required to be signed by
Although the grammar of the keys clause allows for multiple keys, only a
single key per server is currently supported.
6.2.15 trusted-keys Statement Grammar
string number number number string ;
[ string number number number string ; [...]]
6.2.16 trusted-keys Statement Definition and Usage
The trusted-keys statement defines DNSSEC security roots. See DNSSEC for a
description. A security root is defined when the public key for a
non-authoritative zone is known, but cannot be securely obtained through
DNS, either because it is the DNS root zone or its parent zone is unsigned.
Once a key has been configured as a trusted key, it is treated as if it had
been validated and proven secure. The resolver attempts DNSSEC validation on
all DNS data in subdomains of a security root.
The trusted-keys statement can contain multiple key entries, each consisting
of the key's domain name, flags, protocol, algorithm, and the base-64
representation of the key data.
6.2.17 view Statement Grammar
match_clients { address_match_list } ;
6.2.18 view Statement Definition and Usage
The view statement is a powerful new feature of BIND 9 that lets a name
server answer a DNS query differently depending on who is asking. It is
particularly useful for implementing split DNS setups without having to run
Each view statement defines a view of the DNS namespace that will be seen by
those clients whose IP addresses match the address_match_list of the view's
match-clients clause. The order of the view statements is significant--a
client query will be resolved in the context of the first view whose
match-clients list matches the client's IP address.
Zones defined within a view statement will be only be accessible to clients
that match the view . By defining a zone of the same name in multiple views,
different zone data can be given to different clients, for example,
"internal" and "external" clients in a split DNS setup.
Many of the options given in the options statement can also be used within a
view statement, and then apply only when resolving queries with that view.
When no a view-specific value is given, the value in the options statement
is used as a default. Also, zone options can have default values specified
in the view statement; these view-specific defaults take precedence over
those in the options statement.
Views are class specific. If no class is given, class IN is assumed.
If there are no view statements in the config file, a default view that
matches any client is automatically created in class IN, and any zone
statements specified on the top level of the configuration file are
considered to be part of this default view. If any explicit view statements
are present, all zone statements must occur inside view statements.
A zone statement of type hint for the root zone (` . ') does not strictly
define a zone. Therefore, it should not be included in a view statement.
Here is an example of a typical split DNS setup implemented using view
// This should match our internal networks.
match-clients { 10.0.0.0/8; };
// Provide recursive service to internal clients only.
// including addresses of internal hosts.
// Refuse recursive service to external clients.
// containing only publicly accessible hosts.
6.2.19 zone Statement Grammar
zone zone name [class] [{
type ( master|slave|hint|stub|forward ) ;
[ allow-query { address_match_list } ; ]
[ allow-transfer { address_match_list } ; ]
[ allow-update { address_match_list } ; ]
[ update-policy { update_policy_rule[...] } ; ]
[ allow-update-forwarding { address_match_list } ; ]
[ also-notify { [ ip_addr ; [ip_addr ; [...]]] } ; ]
[ check-names (warn|fail|ignore) ; ]
[ dialup true_or_false ; ]
[ forward (only|first) ; ]
[ forwarders { [ ip_addr ; [ ip_addr ; [...]]] } ; ]
[ ixfr-tmp-file string ; ]
[ maintain-ixfr-base true_or_false ; ]
[ masters [port number] { ip_addr ; [ip_addr ; [...]] } ; ]
[ max-ixfr-log-size number ; ]
[ max-transfer-idle-in number ; ]
[ max-transfer-idle-out number ; ]
[ max-transfer-time-in number ; ]
[ max-transfer-time-out number ; ]
[ notify true_or_false ; ]
[ pubkey number number number string ; ]
[ transfer-source (ip4_addr | *) ; ]
[ transfer-source-v6 (ip6_addr | *) ; ]
[ sig-validity-interval number ; ]}]
6.2.20 zone Statement Definition and Usage
master The server has a master copy of the data for the zone and will be
able to provide authoritative answers for it.
A slave zone is a replica of a master zone. The masters list
specifies one or more IP addresses that the slave contacts to
update its copy of the zone. If a port is specified, the slave
then checks to see if the zone is current and zone transfers will
be done to the port given. If a file is specified, then the
replica will be written to this file whenever the zone is changed,
and reloaded from this file on a server restart. Use of a file is
slave recommended, since it often speeds server start-up and eliminates
a needless waste of bandwidth. Note that for large numbers (in the
tens or hundreds of thousands) of zones per server, it is best to
use a two level naming scheme for zone file names. For example, a
slave server for the zone
example.com might place the zone
contents into a file called
name. (Most operating systems behave very slowly if you put 100K
files into a single directory.)
A stub zone is similar to a slave zone, except that it replicates
only the NS records of a master zone instead of the entire zone.
Stub zones are not a standard part of the DNS; they are a
peculiarity of BIND 4 and BIND 8 that relies heavily on the
particular way the zone data is structured in those servers.
BIND 9 attempts to emulate the BIND 4/8 stub zone feature for
backwards compatibility, but we do not recommend its use in new
In BIND 4/8, zone transfers of a parent zone included the NS
records from stub children of that zone. This meant that, in some
cases, users could get away with configuring child stubs only in
the master server for the parent zone. BIND 9 never mixes together
zone data from different zones in this way. Therefore, if a BIND 9
master serving a parent zone has child stub zones configured, all
the slave servers for the parent zone also need to have the same
child stub zones configured..
A "forward zone" is a way to configure forwarding on a per-domain
basis. A zone statement of type forward can contain a forward
and/or forwarders statement, which will apply to queries within
the domain given by the zone name. If no forwarders statement is
present or an empty list for forwarders is given, then no
forward forwarding will be done for the domain, cancelling the effects of
any forwarders in the options statement. Thus if you want to use
this type of zone to change the behavior of the global forward
option (that is, "forward first to", then "forward only", or vice
versa, but want to use the same servers as set globally) you need
to respecify the global forwarders. Domain-specific forwarding is
not yet implemented in BIND 9.
The initial set of root nameservers is specified using a "hint
zone". When the server starts up, it uses the root hints to find a
hint root nameserver and get the most recent list of root nameservers.
If no hint zone is specified for class IN, the server users a
compiled-in default set of root servers hints. Classes other than
IN have no built-in defaults hints.
The zone's name may optionally be followed by a class. If a class is not
specified, class IN (for Internet ), is assumed. This is correct for the
The hesiod class is named for an information service from MIT's Project
Athena. It is used to share information about various systems databases,
such as users, groups, printers and so on. The keyword HS is a synonym for
Another MIT development is CHAOSnet, a LAN protocol created in the
mid-1970s. Zone data for it can be specified with the CHAOS class.
allow-query See the description of allow-query under Access
allow-transfer See the description of allow-transfer under Access
Specifies which hosts are allowed to submit
allow-update Dynamic DNS updates for master zones. The default
is to deny updates from all hosts.
update-policy Specifies a "Simple Secure Update" policy. See
description in Dynamic Update Policies .
Specifies which hosts are allowed to submit
Dynamic DNS updates to slave zones to be forwarded
allow-update-forwarding to the master. The default is to deny update
forwarding from all hosts. Update forwarding is
Only meaningful if notify is active for this zone.
The set of machines that will receive a DNS NOTIFY
message for this zone is made up of all the listed
also-notify nameservers (other than the primary master) for
the zone plus any IP addresses specified with
also-notify is not meaningful for stub zones. The
default is the empty list .
check-names See Name Checking .
Not yet implemented in BIND 9.
See the description of dialup under Boolean
Not yet implemented in BIND 9.
Only meaningful if the zone has a forwarders list.
The only value causes the lookup to fail after
forward trying the forwarders and getting no answer, while
first would allow a normal lookup to be tried.
Not yet implemented in BIND 9.
Used to override the list of global forwarders. If
it is not specified in a zone of type forward , no
forwarders forwarding is done for the zone; the global
Not yet implemented in BIND 9.
Was used in BIND 8 to specify the name of the
transaction log (journal) file for dynamic update
ixfr-base and IXFR. BIND 9 ignores the option and constructs
the name of the journal file by appending ". jnl "
to the name of the zone file.
max-transfer-time-in See the description of
max-transfer-time-in under Zone Transfers .
max-transfer-idle-in See the description of
max-transfer-idle-in under Zone Transfers .
max-transfer-time-out See the description of
max-transfer-time-out under Zone Transfers .
max-transfer-idle-out See the description of
max-transfer-idle-out under Zone Transfers .
notify See the description of notify under Boolean
In BIND 8, this option was intended for specifying
a public zone key for verification of signatures
pubkey in DNSSEC signed zones when they are loaded from
disk. BIND 9 does not verify signatures on loading
sig-validity-interval See the description of sig-validity-interval in
Determines which local address will be bound to
the IPv4 TCP connection used to fetch this zone.
If not set, it defaults to a system controlled
transfer-source value which will usually be the address of the
interface "closest to" the remote end. This
address must appear in the remote end's
allow-transfer option for this zone if one is
transfer-source-v6 Similar to transfer-source, but for zone transfers
6.2.20.4 Dynamic Update Policies
BIND 9 supports two alternative methods of granting clients the right to
perform dynamic updates to a zone, configured by the allow-update and
update-policy option, respectively.
The allow-update clause works the same way as in previous versions of BIND.
It grants given clients the permission to update any record of any name in
The update-policy clause is new in BIND 9 and allows more fine-grained
control over what updates are allowed. A set of rules is specified, where
each rule either grants or denies permissions for one or more names to be
updated by one or more identities. If the dynamic update request message is
signed (that is, it includes either a TSIG or SIG(0) record), the identity
of the signer can be determined.
Rules are specified in the update-policy zone option, and are only
meaningful for master zones. When the update-policy statement is present, it
is a configuration error for the allow-update statement to be present. The
update-policy statement only examines the signer of a message; the source
This is how a rule definition looks:
( grant | deny ) identity nametype name [ types ]
Each rule grants or denies privileges. Once a messages has successfully
matched a rule, the operation is immediately granted or denied and no
further rules are examined. A rule is matched when the signer matches the
identity field, the name matches the name field, and the type is specified
The identity field specifies a name or a wildcard name. The nametype field
has 4 values: name , subdomain , wildcard , and self .
name Matches when the updated name is the same as the name in the
subdomain Matches when the updated name is a subdomain of the name in the
wildcard Matches when the updated name is a valid expansion of the
wildcard name in the name field.
self Matches when the updated name is the same as the message signer.
The name field is ignored.
If no types are specified, the rule matches all types except SIG, NS, SOA,
and NXT. Types may be specified by name, including "ANY" (ANY matches all
types except NXT, which can never be updated).
6.3.1 Types of Resource Records and When to Use Them
This section, largely borrowed from RFC 1034, describes the concept of a
Resource Record (RR) and explains when each is used. Since the publication
of RFC 1034, several new RRs have been identified and implemented in the
DNS. These are also included.
A domain name identifies a node. Each node has a set of resource
information, which may be empty. The set of resource information associated
with a particular name is composed of separate RRs. The order of RRs in a
set is not significant and need not be preserved by nameservers, resolvers,
or other parts of the DNS. However, sorting of multiple RRs is permitted for
optimization purposes, for example, to specify that a particular nearby
server be tried first. See The sortlist Statement and RRset Ordering for
The components of a Resource Record are
owner namethe domain name where the RR is found.
type an encoded 16 bit value that specifies the type of the resource
in this resource record. Types refer to abstract resources.
the time to live of the RR. This field is a 32 bit integer in
TTL units of seconds, and is primarily used by resolvers when they
cache RRs. The TTL describes how long a RR can be cached before
class an encoded 16 bit value that identifies a protocol family or
RDATA the type and sometimes class-dependent data that describes the
The following are types of valid RRs (some of these listed, although not
obsolete, are experimental (x) or historical (h) and no longer in general
AAAA Obsolete format of IPv6 address
AFSDB(x) location of AFS database servers. Experimental.
CNAMEidentifies the canonical name of an alias.
DNAMEfor delegation of reverse addresses. Replaces the domain name
specified with another name to be looked up. Described in RFC 2672.
HINFOidentifies the CPU and OS used by a host.
ISDN (x) representation of ISDN addresses. Experimental.
KEY stores a public key associated with a DNS name.
LOC (x) for storing GPS info. See RFC 1876. Experimental.
MX identifies a mail exchange for the domain. See RFC 974 for details.
NS the authoritative nameserver for the domain.
used in DNSSEC to securely indicate that RRs with an owner name in a
NXT certain name interval do not exist in a zone and indicate what RR
types are present for an existing name. See RFC 2535 for details.
PTR a pointer to another part of the domain name space.
RP (x) information on persons responsible for the domain. Experimental.
RT (x) route-through binding for hosts that do not have their own direct
wide area network addresses. Experimental.
SIG ("signature") contains data authenticated in the secure DNS. See RFC
SOA identifies the start of a zone of authority.
SRV information about well known network services (replaces WKS).
WKS (h) information about which well known network services, such as
SMTP, that a domain supports. Historical, replaced by newer RR SRV.
X25 (x) representation of X.25 network addresses. Experimental.
The following classes of resource records are currently valid in the DNS:
For information about other, older classes of RRs, see Classes of Resource
RDATA is the type-dependent or class-dependent data that describes the
A for the IN class, a 32 bit IP address.
A6 maps a domain name to an IPv6 address, with a provision for
indirection for leading "prefix" bits.
provides alternate naming to an entire subtree of the domain name
DNAMEspace, rather than to a single node. It causes some suffix of a
queried name to be substituted with a name from the DNAME record's
MX a 16 bit preference value (lower is better) followed by a host name
willing to act as a mail exchange for the owner domain.
NS a fully qualified domain name.
PTR a fully qualified domain name.
The owner name is often implicit, rather than forming an integral part of
the RR. For example, many nameservers internally form tree or hash
structures for the name space, and chain RRs off nodes. The remaining RR
parts are the fixed header (type, class, TTL) which is consistent for all
RRs, and a variable part (RDATA) that fits the needs of the resource being
The meaning of the TTL field is a time limit on how long an RR can be kept
in a cache. This limit does not apply to authoritative data in zones; it is
also timed out, but by the refreshing policies for the zone. The TTL is
assigned by the administrator for the zone where the data originates. While
short TTLs can be used to minimize caching, and a zero TTL prohibits
caching, the realities of Internet performance suggest that these times
should be on the order of days for the typical host. If a change can be
anticipated, the TTL can be reduced prior to the change to minimize
inconsistency during the change, and then increased back to its former value
The data in the RDATA section of RRs is carried as a combination of binary
strings and domain names. The domain names are frequently used as "pointers"
to other data in the DNS.
6.3.1.2 Textual expression of RRs
RRs are represented in binary form in the packets of the DNS protocol, and
are usually represented in highly encoded form when stored in a nameserver
or resolver. In the examples provided in RFC 1034, a style similar to that
used in master files was employed in order to show the contents of RRs. In
this format, most RRs are shown on a single line, although continuation
lines are possible using parentheses.
The start of the line gives the owner of the RR. If a line begins with a
blank, then the owner is assumed to be the same as that of the previous RR.
Blank lines are often included for readability.
Following the owner, we list the TTL, type, and class of the RR. Class and
type use the mnemonics defined above, and TTL is an integer before the type
field. In order to avoid ambiguity in parsing, type and class mnemonics are
disjoint, TTLs are integers, and the type mnemonic is always last. The IN
class and TTL values are often omitted from examples in the interests of
The resource data or RDATA section of the RR are given using knowledge of
the typical representation for the data.
For example, we might show the RRs carried in a message as:
The MX RRs have an RDATA section which consists of a 16 bit number followed
by a domain name. The address RRs use a standard IP address format to
contain a 32 bit internet address.
This example shows six RRs, with two RRs at each of three domain names.
This example shows two addresses for
XX.LCS.MIT.EDU , each of a different
6.3.2 Discussion of MX Records
As described above, domain servers store information as a series of resource
records, each of which contains a particular piece of information about a
given domain name (which is usually, but not always, a host). The simplest
way to think of a RR is as a typed pair of datum, a domain name matched with
relevant data, and stored with some additional type information to help
systems determine when the RR is relevant.
MX records are used to control delivery of email. The data specified in the
record is a priority and a domain name. The priority controls the order in
which email delivery is attempted, with the lowest number first. If two
priorities are the same, a server is chosen randomly. If no servers at a
given priority are responding, the mail transport agent will fall back to
the next largest priority. Priority numbers do not have any absolute meaning
- they are relevant only respective to other MX records for that domain
name. The domain name given is the machine to which the mail will be
delivered. It must have an associated A record--a CNAME is not sufficient.
For a given domain, if there is both a CNAME record and an MX record, the MX
record is in error, and will be ignored. Instead, the mail will be delivered
to the server specified in the MX record pointed to by the CNAME.
(in any order), and if neither of those succeed, delivery to
mail.backup.orgThe time to live of the RR field is a 32 bit integer represented in units of
seconds, and is primarily used by resolvers when they cache RRs. The TTL
describes how long a RR can be cached before it should be discarded. The
following three types of TTL are currently used in a zone file.
The last field in the SOA is the negative caching TTL. This
controls how long other servers will cache no-such-domain
SOA (NXDOMAIN) responses from you.
The maximum time for negative caching is 3 hours (3h).
$TTL The $TTL directive at the top of the zone file (before the SOA)
gives a default TTL for every RR without a specific TTL set.
RR TTLsEach RR can have a TTL as the second field in the RR, which will
control how long other servers can cache the it.
All of these TTLs default to units of seconds, though units can be
explicitly specified, for example, 1h30m .
6.3.4 Inverse Mapping in IPv4
Reverse name resolution (that is, translation from IP address to name) is
achieved by means of the
in-addr.arpa domain and PTR records. Entries in the
in-addr.arpa domain are made in least-to-most significant order, read left
to right. This is the opposite order to the way IP addresses are usually
written. Thus, a machine with an IP address of 10.1.2.3 would have a
data field is the name of the machine or, optionally, multiple PTR records
if the machine has more than one name. For example, in the
example.com(Note: The $ORIGIN lines in the examples are for providing context to the
examples only--they do not necessarily appear in the actual usage. They are
only used here to indicate that the example is relative to the listed
6.3.5 Other Zone File Directives
The Master File Format was initially defined in RFC 1035 and has
subsequently been extended. While the Master File Format itself is class
independent all records in a Master File must be of the same class.
Master File Directives include $ORIGIN , $INCLUDE , and $TTL.
6.3.5.1 The $ORIGIN Directive
Syntax: $ORIGIN < domain-name > [ < comment > ]
$ORIGIN sets the domain name that will be appended to any unqualified
records. When a zone is first read in there is an implicit $ORIGIN <
zone-name > . The current $ORIGIN is appended to the domain specified in the
$ORIGIN argument if it is not absolute.
6.3.5.2 The $INCLUDE Directive
Syntax: $INCLUDE < filename > [ < origin > ] [ < comment > ]
Read and process the file filename as if it were included into the file at
this point. If origin is specified the file is processed with $ORIGIN set to
that value, otherwise the current $ORIGIN is used.
NOTE: The behavior when origin is specified differs from that described in
RFC 1035. The origin and current domain revert to the values they were prior
to the $INCLUDE once the file has been read.
6.3.5.3 The $TTL Directive
Syntax: $TTL < default-ttl > [ < comment > ]
Set the default Time To Live (TTL) for subsequent records with undefined
TTLs. Valid TTLs are of the range 0-2147483647 seconds.
$TTL is defined in RFC 2308.
6.3.6 BIND Master File Extension: the $GENERATE Directive
Syntax: $GENERATE < range > < lhs > < type > < rhs > [ < comment > ]
$GENERATE is used to create a series of resource records that only differ
from each other by an iterator. $GENERATE can be used to easily generate the
sets of records required to support sub /24 reverse delegations described in
$GENERATE 1-2 0 NS SERVER$.EXAMPLE.
$GENERATE 1-127 $ CNAME $.0
range first form is used then step is set to 1. All of start, stop and
lhs describes the owner name of the resource records to be created.
Any single $ symbols within the lhs side are replaced by the
lhs iterator value. To get a $ in the output use a double $ ,
e.g. $$ .
If the lhs is not absolute, the current $ORIGIN is appended to the
type At present the only supported types are PTR, CNAME and NS.
rhs rhs is a domain name. It is processed similarly to lhs.
The $GENERATE directive is a BIND extension and not part of the standard
zone file format. It is not yet implemented in BIND 9.
Certain UNIX signals cause the name server to take specific actions, as
described in the following table. These signals can be sent using the kill
SIGHUP Causes the server to read
named.conf and reload the database.
SIGTERM Causes the server to clean up and exit.
SIGINT Causes the server to clean up and exit.
SIGQUIT Causes the server to clean up and exit.
------------------------------------------------------------------------
Section 7. BIND 9 Security Considerations
Access Control Lists (ACLs), are address match lists that you can set up and
nickname for future use in allow-query , allow-recursion , blackhole ,
Using ACLs allows you to have finer control over who can access your
nameserver, without cluttering up your config files with huge lists of IP
It is a good idea to use ACLs, and to control access to your server.
Limiting access to your server by outside parties can help prevent spoofing
and DoS attacks against your server.
Here is an example of how to properly apply ACLs:
// Set up an ACL named "bogusnets" that will block RFC1918 space,
// which is commonly used in spoofing attacks.
acl bogusnets { 0.0.0.0/8; 1.0.0.0/8; 2.0.0.0/8; 192.0.2.0/24; 224.0.0.0/3;
10.0.0.0/8; 172.16.0.0/12; 192.168.0.0/16; };
// Set up an ACL called our-nets. Replace this with the real IP numbers.
allow-query { our-nets; };
allow-recursion { our-nets; };
blackhole { bogusnets; };
This allows recursive queries of the server from the outside unless
recursion has been previously disabled.
For more information on how to use ACLs to protect your server, see the
7.2 chroot and setuid (for UNIX servers)
On UNIX servers, it is possible to run BIND in a chrooted environment (
chroot() ) by specifying the " -t " option. This can help improve system
security by placing BIND in a "sandbox," which will limit the damage done if
Another useful feature in the UNIX version of BIND is the ability to run the
daemon as a nonprivileged user ( -u user ). We suggest running as a
nonprivileged user when using the chroot feature.
Here is an example command line to load BIND in a chroot() sandbox,
7.2.1 The chroot Environment
In order for a chroot() environment to work properly in a particular
directory (for example,
/var/named ), you will need to set up an environment
that includes everything BIND needs to run. From BIND's point of view,
library directories and files that BIND needs to run on your system. Please
consult your operating system's instructions if you need help figuring out
which library files you need to copy over to the chroot() sandbox.
If you are running an operating system that supports static binaries, you
can also compile BIND statically and avoid the need to copy system libraries
over to your chroot() sandbox.
7.2.2 Using the setuid Function
Prior to running the named daemon, use the touch utility (to change file
access and modification times) or the chown utility (to set the user id
and/or group id) on files to which you want BIND to write.
Access to the dynamic update facility should be strictly limited. In earlier
versions of BIND the only way to do this was based on the IP address of the
host requesting the update. BIND 9BIND 9 also supports authenticating
updates cryptographically by means of transaction signatures (TSIG). The use
of TSIG is strongly recommended.
Some sites choose to keep all dynamically updated DNS data in a subdomain
and delegate that subdomain to a separate zone. This way, the top-level zone
containing critical data such as the IP addresses of public web and mail
servers need not allow dynamic update at all.
------------------------------------------------------------------------
Section 8. Troubleshooting
8.1.1 It's not working; how can I figure out what's wrong?
The best solution to solving installation and configuration issues is to
take preventative measures by setting up logging files beforehand. (See the
sample configurations) in Section 3. The log files provide a source of hints
and information that can be used to figure out what went wrong and how to
8.2 Incrementing and Changing the Serial Number
Zone serial numbers are just numbers--they aren't date related. A lot of
people set them to a number that represents a date, usually of the form
YYYYMMDDRR. A number of people have been testing these numbers for Y2K
compliance and have set the number to the year 2000 to see if it will work.
They then try to restore the old serial number. This will cause problems
because serial numbers are used to indicate that a zone has been updated. If
the serial number on the slave server is lower than the serial number on the
master, the slave server will attempt to update its copy of the zone.
Setting the serial number to a lower number on the master server than the
slave server means that the slave will not perform updates to its copy of
The solution to this is to add 2147483647 (2^31-1) to the number, reload the
zone and make sure all slaves have updated to the new zone serial number,
then reset the number to what you want it to be, and reload the zone again.
8.3 Where Can I Get Help?
The Internet Software Consortium (ISC) offers a wide range of support and
service agreements for BIND and DHCP servers. Four levels of premium support
are available and each level includes support for all ISC programs,
significant discounts on products and training, and a recognized priority on
bug fixes and non-funded feature requests. In addition, ISC offers a
standard support agreement package which includes services ranging from bug
fix announcements to remote support. It also includes training in BIND and
To discuss arrangements for support, contact info@isc.org or visit the ISC
------------------------------------------------------------------------
A Brief History of the DNS and BIND
Although the "official" beginning of the Domain Name System occurred in 1984
with the publication of RFC 920, the core of the new system was described in
1983 in RFCs 882 and 883. From 1984 to 1987, the ARPAnet (the precursor to
today's Internet) became a testbed of experimentation for developing the new
environment. New RFCs were written and published in 1987 that modified the
original documents to incorporate improvements based on the working model.
RFC 1034, "Domain Names-Concepts and Facilities," and RFC 1035, "Domain
Names-Implementation and Specification" were published and became the
standards upon which all DNS implementations are built.
The first working domain name server, called "Jeeves," was written in
1983-84 by Paul Mockapetris for operation on DEC Tops-20 machines located at
the University of Southern California's Information Sciences Institute
(USC-ISI) and SRI International's Network Information Center (SRI-NIC). A
DNS server for Unix machines, the Berkeley Internet Name Domain (BIND)
package, was written soon after by a group of graduate students at the
University of California at Berkeley under a grant from the US Defense
Advanced Research Projects Administration (DARPA). Versions of BIND through
4.8.3 were maintained by the Computer Systems Research Group (CSRG) at UC
Berkeley. Douglas Terry, Mark Painter, David Riggle and Songnian Zhou made
up the initial BIND project team. After that, additional work on the
software package was done by Ralph Campbell. Kevin Dunlap, a Digital
Equipment Corporation employee on loan to the CSRG, worked on BIND for 2
years, from 1985 to 1987. Many other people also contributed to BIND
development during that time: Doug Kingston, Craig Partridge, Smoot
Carl-Mitchell, Mike Muuss, Jim Bloom and Mike Schwartz. BIND maintenance was
subsequently handled by Mike Karels and O. Kure.
BIND versions 4.9 and 4.9.1 were released by Digital Equipment Corporation
(now Compaq Computer Corporation). Paul Vixie, then a DEC employee, became
BIND's primary caretaker. Paul was assisted by Phil Almquist, Robert Elz,
Alan Barrett, Paul Albitz, Bryan Beecher, Andrew Partan, Andy Cherenson, Tom
Limoncelli, Berthold Paffrath, Fuat Baran, Anant Kumar, Art Harkin, Win
Treese, Don Lewis, Christophe Wolfhugel, and others.
BIND Version 4.9.2 was sponsored by Vixie Enterprises. Paul Vixie became
BIND versions from 4.9.3 onward have been developed and maintained by the
Internet Software Consortium with support being provided by ISC's sponsors.
production-ready version of BIND version 8 in May 1997.
BIND development work is made possible today by the sponsorship of several
corporations, and by the tireless work efforts of numerous individuals.
Historical DNS Information
Classes of Resource Records
The hesiod class is an information service developed by MIT's Project
Athena. It is used to share information about various systems databases,
such as users, groups, printers and so on. The keyword hs is a synonym for
The chaos class is used to specify zone data for the MIT-developed CHAOSnet,
a LAN protocol created in the mid-1970s.
General DNS Reference Information
IPv6 addresses are 128-bit identifiers for interfaces and sets of interfaces
which were introduced in the DNS to facilitate scalable Internet routing.
There are three types of addresses: Unicast , an identifier for a single
interface; Anycast , an identifier for a set of interfaces; and Multicast ,
an identifier for a set of interfaces. Here we describe the global Unicast
address scheme. For more information, see RFC 2374.
The aggregatable global Unicast address format is as follows:
FP TLA ID RES NLA ID SLA ID Interface ID
<------ Interface Identifier
TLA ID = Top-Level Aggregation Identifier
RES = Reserved for future use
NLA ID = Next-Level Aggregation Identifier
SLA ID = Site-Level Aggregation Identifier
INTERFACE ID= Interface Identifier
The Public Topology is provided by the upstream provider or ISP, and
(roughly) corresponds to the IPv4 network section of the address range. The
Site Topology is where you can subnet this space, much the same as
subnetting an IPv4 class A or B network into class Cs. The Interface
Identifier is the address of an individual interface on a given network.
(With IPv6, addresses belong to interfaces rather than machines.)
The subnetting capability of IPv6 is much more flexible than that of IPv4:
subnetting can now be carried out on bit boundaries, in much the same way as
Classless InterDomain Routing (CIDR).
The internal structure of the Public Topology for an A6 global unicast
A 3 bit FP (Format Prefix) of 001 indicates this is a global Unicast
address. FP lengths for other types of addresses may vary.
13 TLA (Top Level Aggregator) bits give the prefix of your top-level IP
24 bits for Next Level Aggregators. This allows organizations with a TLA to
hand out portions of their IP space to client organizations, so that the
client can then split up the network further by filling in more NLA bits,
and hand out IPv6 prefixes to their clients, and so forth.
There is no particular structure for the Site topology section.
Organizations can allocate these bits in any way they desire, in the same
way as they would subnet an IPv4 class A (8-bit prefix) network.
The Interface Identifier must be unique on that network. On ethernet
networks, one way to ensure this is to set the address to the first three
bytes of the hardware address, "FFFE", then the last three bytes of the
hardware address. The lowest significant bit of the first byte should then
be complemented. Addresses are written as 32-bit blocks separated with a
colon, and leading zeros of a block may be omitted, for example:
3ffe:8050:201:9:a00:20ff:fe81:2b32
IPv6 address specifications are likely to contain long strings of zeros, so
the architects have included a shorthand for specifying them. The double
colon (`::') indicates the longest possible string of zeros that can fit,
and can be used only once in an address.
Bibliography (and Suggested Reading)
Request for Comments (RFCs)
Specification documents for the Internet protocol suite, including the DNS,
are published as part of the Request for Comments (RFCs) series of technical
notes. The standards themselves are defined by the Internet Engineering Task
Force (IETF) and the Internet Engineering Steering Group (IESG). RFCs can be
obtained online via FTP at
RFC974. Partridge, C. Mail Routing and the Domain System . January 1986.
RFC1034. Mockapetris,
P.V. Domain Names - Concepts and Facilities .
P.V.RFC1035. Mockapetris, P. V. Domain Names - Implementation and Specification
RFC2181. Elz, R., R. Bush. Clarifications to the DNS Specification . July
RFC2308. Andrews, M. Negative Caching of DNS Queries . March 1998.
RFC1995. Ohta, M. Incremental Zone Transfer in DNS . August 1996.
RFC1996. Vixie, P. A Mechanism for Prompt Notification of Zone Changes .
RFC2136. Vixie, P., S. Thomson, Y. Rekhter, J. Bound. Dynamic Updates in the
Domain Name System . April 1997.
RFC2845. Vixie, P., O. Gudmundsson, D. Eastlake 3rd, B. Wellington. Secret
Key Transaction Authentication for DNS (TSIG) . May 2000.
Proposed Standards Still Under Development
Note: the following list of RFCs are undergoing major revision by the IETF.
RFC1886. Thomson, S., C. Huitema. DNS Extensions to support IP version 6 .
RFC2065. Eastlake, 3rd, D., C. Kaufman. Domain Name System Security
Extensions . January 1997.
RFC2137. Eastlake, 3rd, D. Secure Domain Name System Dynamic Update . April
Other Important RFCs About DNS Implementation
RFC1535. Gavron, E. A Security Problem and Proposed Correction With Widely
Deployed DNS Software. October 1993.
RFC1536. Kumar, A., J. Postel, C. Neuman, P. Danzig, S. Miller. Common DNS
Implementation Errors and Suggested Fixes . October 1993.
RFC1982. Elz, R., R. Bush. Serial Number Arithmetic . August 1996.
RFC1183. Everhart,
C.F., L. A. Mamakos, R. Ullmann, P. Mockapetris. New DNS
RR Definitions . October 1990.
RFC1706. Manning, B., R. Colella. DNS NSAP Resource Records . October 1994.
RFC2168. Daniel, R., M. Mealling. Resolution of Uniform Resource Identifiers
using the Domain Name System. June 1997.
RFC1876. Davis, C., P. Vixie, T. Goodwin, I. Dickinson. A Means for
Expressing Location Information in the Domain Name System . January 1996.
RFC2052. Gulbrandsen, A., P. Vixie. A DNS RR for Specifying the Location of
RFC2163. Allocchio, A. U sing the Internet DNS to Distribute MIXER
Conformant Global Address Mapping . January 1998.
RFC2230. Atkinson, R. Key Exchange Delegation Record for the DNS . October
RFC1101. Mockapetris, P. V. DNS Encoding of Network Names and Other Types .
RFC1123. Braden, R. Requirements for Internet Hosts - Application and
RFC1591. Postel, J. D omain Name System Structure and Delegation . March
RFC2317. Eidnes, H., G. de Groot, P. Vixie. Classless
IN-ADDR.ARPARFC1537. Beertema, P. Common DNS Data File Configuration Errors . October
RFC1912. Barr, D. Common DNS Operational and Configuration Errors . February
RFC1912. Barr, D. Common DNS Operational and Configuration Errors . February
RFC2010. Manning, B., P. Vixie. Operational Criteria for Root Name Servers.
RFC2219. Hamilton, M., R. Wright. Use of DNS Aliases for Network Services.
Note: the following list of RFCs, although DNS-related, are not concerned
with implementing software.
RFC1464. Rosenbaum, R. Using the Domain Name System To Store Arbitrary
String Attributes . May 1993.
RFC1713. Romao, A. Tools for DNS Debugging . November 1994.
RFC1794. Brisco, T. DNS Support for Load Balancing . April 1995.
RFC2240. Vaughan, O. A Legal Basis for Domain Name Allocation .
RFC2345. Klensin, J., T. Wolf, G. Oglesby. Domain Names and Company Name
RFC2352. Vaughan, O. A Convention For Using Legal Names as Domain Names .
Obsolete and Unimplemented Experimental RRs
RFC1712. Farrell, C., M. Schulze, S. Pleitner, D. Baldoni. DNS Encoding of
Geographical Location . November 1994.
Internet Drafts (IDs) are rough-draft working documents of the Internet
Engineering Task Force. They are, in essence, RFCs in the preliminary stages
of development. Implementors are cautioned not to regard IDs as archival,
and they should not be quoted or cited in any formal documents unless
accompanied by the disclaimer that they are "works in progress." IDs have a
lifespan of six months after which they are deleted unless updated by their
Albitz, Paul and Cricket Liu. 1998. DNS and BIND . Sebastopol, CA: O'Reilly
------------------------------------------------------------------------