dhcpd.conf, dhcpd-dhcpv6.conf

Updated: April 19, 2023

DHCP configuration files

Name:

Description:

The dhcpd.conf file contains configuration information for dhcpd, the Internet Systems Consortium DHCP Server.

The dhcpd.conf file is a free-form ASCII text file. It's parsed by the recursive-descent parser built into dhcpd. The file may contain extra tabs and newlines for formatting purposes. Keywords in the file are case-insensitive. Comments may be placed anywhere within the file (except within quotes). Comments begin with the # character and end at the end of the line.

The file essentially consists of a list of statements. Statements fall into two broad categories: parameters and declarations.

Parameter statements either say how to do something (e.g., how long a lease to offer), whether to do something (e.g., should dhcpd provide addresses to unknown clients), or what parameters to provide to the client (e.g., use gateway 220.177.244.7).

Declarations are used to describe the topology of the network, to describe clients on the network, to provide addresses that can be assigned to clients, or to apply a group of parameters to a group of declarations. In any group of parameters and declarations, all parameters must be specified before any declarations that depend on those parameters may be specified.

Declarations about network topology include the shared-network and the subnet declarations. If clients on a subnet are to be assigned addresses dynamically, a range declaration must appear within the subnet declaration. For clients with statically assigned addresses, or for installations where only known clients will be served, each such client must have a host declaration. If parameters are to be applied to a group of declarations which aren't related strictly on a per-subnet basis, the group declaration can be used.

For every subnet which will be served, and for every subnet to which the DHCP server is connected, there must be one subnet declaration, which tells dhcpd how to recognize that an address is on that subnet. A subnet declaration is required for each subnet even if no addresses will be dynamically allocated on that subnet.

Some installations have physical networks on which more than one IP subnet operates. For example, if there is a site-wide requirement that 8-bit subnet masks be used, but a department with a single physical ethernet network expands to the point where it has more than 254 nodes, it may be necessary to run two 8-bit subnets on the same ethernet until such time as a new physical network can be added. In this case, the subnet declarations for these two networks must be enclosed in a shared-network declaration.

Note that even when the shared-network declaration is absent, an empty one is created by the server to contain the subnet (and any scoped parameters included in the subnet). For practical purposes, this means that “stateless” DHCP clients, which aren't tied to addresses (and therefore subnets) will receive the same configuration as stateful ones.

Some sites may have departments which have clients on more than one subnet, but it may be desirable to offer those clients a uniform set of parameters which are different than what would be offered to clients from other departments on the same subnet. For clients which will be declared explicitly with host declarations, these declarations can be enclosed in a group declaration along with the parameters which are common to that department. For clients whose addresses will be dynamically assigned, class declarations and conditional declarations may be used to group parameter assignments based on information the client sends.

When a client is to be booted, its boot parameters are determined by consulting that client's host declaration (if any), and then consulting any class declarations matching the client, followed by the pool, subnet and shared-network declarations for the IP address assigned to the client. Each of these declarations itself appears within a lexical scope, and all declarations at less specific lexical scopes are also consulted for client option declarations. Scopes are never considered twice, and if parameters are declared in more than one scope, the parameter declared in the most specific scope is the one that is used.

When dhcpd tries to find a host declaration for a client, it first looks for a host declaration which has a fixed-address declaration that lists an IP address that is valid for the subnet or shared network on which the client is booting. If it doesn't find any such entry, it tries to find an entry which has no fixed-address declaration.

Examples

A typical dhcpd.conf file will look something like this:

global parameters...

subnet 204.254.239.0 netmask 255.255.255.224 {
  subnet-specific parameters...
  range 204.254.239.10 204.254.239.30;
}

subnet 204.254.239.32 netmask 255.255.255.224 {
  subnet-specific parameters...
  range 204.254.239.42 204.254.239.62;
}

subnet 204.254.239.64 netmask 255.255.255.224 {
  subnet-specific parameters...
  range 204.254.239.74 204.254.239.94;
}

group {
  group-specific parameters...
  host zappo.test.isc.org {
    host-specific parameters...
  }
  host beppo.test.isc.org {
    host-specific parameters...
  }
  host harpo.test.isc.org {
    host-specific parameters...
  }
}

Notice that at the beginning of the file, there's a place for global parameters. These might be things such as the organization's domain name, the addresses of the name servers (if they are common to the entire organization), and so on. So, for example:

option domain-name "isc.org";
option domain-name-servers ns1.isc.org, ns2.isc.org;

As you can see in the above, you can specify host addresses in parameters using their domain names rather than their numeric IP addresses. If a given hostname resolves to more than one IP address (for example, if that host has two ethernet interfaces), then where possible, both addresses are supplied to the client. The most obvious reason for having subnet-specific parameters as shown earlier is that each subnet, of necessity, has its own router. So for the first subnet, for example, there should be something like:

option routers 204.254.239.1;

Note that the address here is specified numerically. This isn't required; if you have a different domain name for each interface on your router, it's perfectly legitimate to use the domain name for that interface instead of the numeric address. However, in many cases there may be only one domain name for all of a router's IP addresses, and it wouldn't be appropriate to use that name here.

In the sample file, there's also a group statement, which provides common parameters for a set of three hosts: zappo, beppo, and harpo. As you can see, these hosts are all in the test.isc.org domain, so it might make sense for a group-specific parameter to override the domain name supplied to these hosts:

option domain-name "test.isc.org";

Also, given the domain they're in, these are probably test machines. If we wanted to test the DHCP leasing mechanism, we might set the lease timeout somewhat shorter than the default:

max-lease-time 120;
default-lease-time 120;

You may have noticed that while some parameters start with the option keyword, some don't. Parameters starting with the option keyword correspond to actual DHCP options, while parameters that don't start with the option keyword either control the behavior of the DHCP server (e.g., how long a lease dhcpd will give out), or specify client parameters that aren't optional in the DHCP protocol (for example, server-name and filename).

In the sample file, each host had host-specific parameters. These could include such things as the hostname option, the name of a file to upload (the filename parameter) and the address of the server from which to upload the file (the next_server parameter). In general, any parameter can appear anywhere that parameters are allowed, and will be applied according to the scope in which the parameter appears.

Imagine that you have a site with a lot of NCD X-Terminals. These terminals come in a variety of models, and you want to specify the boot files for each model. One way to do this would be to have host declarations for each server and group them by model:

group {
  filename "Xncd19r";
  next-server ncd-booter;

  host ncd1 { hardware ethernet 0:c0:c3:49:2b:57; }
  host ncd4 { hardware ethernet 0:c0:c3:80:fc:32; }
  host ncd8 { hardware ethernet 0:c0:c3:22:46:81; }
}

group {
  filename "Xncd19c";
  next-server ncd-booter;

  host ncd2 { hardware ethernet 0:c0:c3:88:2d:81; }
  host ncd3 { hardware ethernet 0:c0:c3:00:14:11; }
}

group {
  filename "XncdHMX";
  next-server ncd-booter;

  host ncd1 { hardware ethernet 0:c0:c3:11:90:23; }
  host ncd4 { hardware ethernet 0:c0:c3:91:a7:8; }
  host ncd8 { hardware ethernet 0:c0:c3:cc:a:8f; }
}

Address pools

The pool declaration can be used to specify a pool of addresses that will be treated differently than another pool of addresses, even on the same network segment or subnet. For example, you may want to provide a large set of addresses that can be assigned to DHCP clients that are registered to your DHCP server, while providing a smaller set of addresses, possibly with short lease times, that are available for unknown clients. If you have a firewall, you may be able to arrange for addresses from one pool to be allowed access to the Internet, while addresses in another pool aren't, thus encouraging users to register their DHCP clients. To do this, you would set up a pair of pool declarations:

subnet 10.0.0.0 netmask 255.255.255.0 {
  option routers 10.0.0.254;

  # Unknown clients get this pool.
  pool {
    option domain-name-servers bogus.example.com;
    max-lease-time 300;
    range 10.0.0.200 10.0.0.253;
    allow unknown-clients;
  }

  # Known clients get this pool.
  pool {
    option domain-name-servers ns1.example.com, ns2.example.com;
    max-lease-time 28800;
    range 10.0.0.5 10.0.0.199;
    deny unknown-clients;
  }
}

It is also possible to set up entirely different subnets for known and unknown clients; address pools exist at the level of shared networks, so address ranges within pool declarations can be on different subnets.

As you can see in the preceding example, pools can have permit lists that control which clients are allowed access to the pool and which aren't. Each entry in a pool's permit list is introduced with the allow or deny keyword. If a pool has a permit list, then only those clients that match specific entries on the permit list will be eligible to be assigned addresses from the pool. If a pool has a deny list, then only those clients that don't match any entries on the deny list will be eligible. If both permit and deny lists exist for a pool, then only clients that match the permit list and don't match the deny list will be allowed access.

Dynamic address allocation

Address allocation is actually only done when a client is in the INIT state and has sent a DHCPDISCOVER message. If the client thinks it has a valid lease and sends a DHCPREQUEST to initiate or renew that lease, the server has only three choices: it can ignore the DHCPREQUEST, send a DHCPNAK to tell the client it should stop using the address, or send a DHCPACK, telling the client to go ahead and use the address for a while.

If the server finds the address the client is requesting, and that address is available to the client, the server will send a DHCPACK. If the address is no longer available, or the client isn't permitted to have it, the server will send a DHCPNAK. If the server knows nothing about the address, it will remain silent, unless the address is incorrect for the network segment to which the client has been attached and the server is authoritative for that network segment, in which case the server will send a DHCPNAK even though it doesn't know about the address.

There may be a host declaration matching the client's identification. If that host declaration contains a fixed-address declaration that lists an IP address that is valid for the network segment to which the client is connected. In this case, the DHCP server will never do dynamic address allocation. In this case, the client is required to take the address specified in the host declaration. If the client sends a DHCPREQUEST for some other address, the server will respond with a DHCPNAK.

When the DHCP server allocates a new address for a client (remember, this happens only if the client has sent a DHCPDISCOVER), it first looks to see if the client already has a valid lease on an IP address, or if there is an old IP address the client had before that hasn't yet been reassigned. In that case, the server will take that address and check it to see if the client is still permitted to use it. If the client is no longer permitted to use it, the lease is freed if the server thought it was still in use; the fact that the client has sent a DHCPDISCOVER proves to the server that the client is no longer using the lease.

If no existing lease is found, or if the client is forbidden to receive the existing lease, then the server will look in the list of address pools for the network segment to which the client is attached for a lease that isn't in use and that the client is permitted to have. It looks through each pool declaration in sequence (all range declarations that appear outside of pool declarations are grouped into a single pool with no permit list). If the permit list for the pool allows the client to be allocated an address from that pool, the pool is examined to see if there is an address available. If so, then the client is tentatively assigned that address. Otherwise, the next pool is tested. If no addresses are found that can be assigned to the client, no response is sent to the client.

If an address is found that the client is permitted to have, and that has never been assigned to any client before, the address is immediately allocated to the client. If the address is available for allocation but has been previously assigned to a different client, the server will keep looking in hopes of finding an address that has never before been assigned to a client.

The DHCP server generates the list of available IP addresses from a hash table. This means that the addresses aren't sorted in any particular order, and so it isn't possible to predict the order in which the DHCP server will allocate IP addresses. Users of previous versions of the ISC DHCP server may have become accustomed to the DHCP server allocating IP addresses in ascending order, but this is no longer possible, and there is no way to configure this behavior with version 3 of the ISC DHCP server.

IP address conflict prevention

The DHCP server checks IP addresses to see if they are in use before allocating them to clients. It does this by sending an ICMP Echo request message to the IP address being allocated. If no ICMP Echo reply is received within a second, the address is assumed to be free. This is done only for leases that have been specified in range statements, and only when the lease is thought by the DHCP server to be free—i.e., the DHCP server or its failover peer hasn't listed the lease as in use.

If a response is received to an ICMP Echo request, the DHCP server assumes that there is a configuration error—the IP address is in use by some host on the network that isn't a DHCP client. It marks the address as abandoned, and will not assign it to clients.

If a DHCP client tries to get an IP address, but none is available, but there are abandoned IP addresses, then the DHCP server will attempt to reclaim an abandoned IP address. It marks one IP address as free, and then does the same ICMP Echo request check described previously. If there is no answer to the ICMP Echo request, the address is assigned to the client.

The DHCP server doesn't cycle through abandoned IP addresses if the first IP address it tries to reclaim is free. Rather, when the next DHCPDISCOVER comes in from the client, it will attempt a new allocation using the same method described here, and will typically try a new IP address.

DHCP Failover

This version of the ISC DHCP server supports the DHCP failover protocol as documented in draft-ietf-dhc-failover-07.txt. This isn't a final protocol document, and we haven't done interoperability testing with other vendors' implementations of this protocol, so you mustn't assume that this implementation conforms to the standard. If you wish to use the failover protocol, make sure that both failover peers are running the same version of the ISC DHCP server.

The failover protocol allows two DHCP servers (and no more than two) to share a common address pool. Each server will have about half of the available IP addresses in the pool at any given time for allocation. If one server fails, the other server will continue to renew leases out of the pool, and will allocate new addresses out of the roughly half of available addresses that it had when communications with the other server were lost.

It is possible during a prolonged failure to tell the remaining server that the other server is down, in which case the remaining server will (over time) reclaim all the addresses the other server had available for allocation, and begin to reuse them. This is called putting the server into the PARTNER-DOWN state.

You can put the server into the PARTNER-DOWN state either by using the omshell command or by stopping the server, editing the last peer state declaration in the lease file, and restarting the server. If you use this last method, be sure to leave the date and time of the start of the state blank:

failover peer name state {
  my state partner-down;
  peer state state at date;
}

When the other server comes back online, it should automatically detect that it has been offline and request a complete update from the server that was running in the PARTNER-DOWN state, and then both servers will resume processing together.

It is possible to get into a dangerous situation: if you put one server into the PARTNER-DOWN state, and then that server goes down, and the other server comes back up, the other server will not know that the first server was in the PARTNER-DOWN state, and may issue addresses previously issued by the other server to different clients, resulting in IP address conflicts. Before putting a server into PARTNER-DOWN state, therefore, make sure that the other server will not restart automatically.

The failover protocol defines a primary server role and a secondary server role. There are some differences in how primaries and secondaries act, but most of the differences simply have to do with providing a way for each peer to behave in the opposite way from the other. So one server must be configured as primary, and the other must be configured as secondary, and it doesn't matter too much which one is which.

Failover startup

When a server starts that hasn't previously communicated with its failover peer, it must establish communications with its failover peer and synchronize with it before it can serve clients. This can happen either because you have just configured your DHCP servers to perform failover for the first time, or because one of your failover servers has failed catastrophically and lost its database.

The initial recovery process is designed to ensure that when one failover peer loses its database and then resynchronizes, any leases that the failed server gave out before it failed will be honored. When the failed server starts up, it notices that it has no saved failover state, and attempts to contact its peer.

When it has established contact, it asks the peer for a complete copy its peer's lease database. The peer then sends its complete database, and sends a message indicating that it is done. The failed server then waits until MCLT has passed, and once MCLT has passed both servers make the transition back into normal operation. This waiting period ensures that any leases the failed server may have given out while out of contact with its partner will have expired.

While the failed server is recovering, its partner remains in the partner-down state, which means that it is serving all clients. The failed server provides no service at all to DHCP clients until it has made the transition into normal operation.

In the case where both servers detect that they have never before communicated with their partner, they both come up in this recovery state and follow the procedure we have just described. In this case, no service will be provided to DHCP clients until MCLT has expired.

Configuring failover

In order to configure failover, you need to write a peer declaration that configures the failover protocol, and you need to write peer references in each pool declaration for which you want to do failover. You don't have to do failover for all pools on a given network segment. You mustn't tell one server it's doing failover on a particular address pool and tell the other it isn't. You mustn't have any common address pools on which you aren't doing failover. A pool declaration that utilizes failover would look like this:

pool {
   failover peer "foo";
   pool-specific parameters
};

The server currently does very little sanity checking, so if you configure it wrong, it will just fail in odd ways. I would recommend therefore that you either do failover or don't do failover, but don't do any mixed pools. Also, use the same master configuration file for both servers, and have a separate file that contains the peer declaration and includes the master file. This will help you to avoid configuration mismatches. As our implementation evolves, this will become less of a problem. A basic sample dhcpd.conf file for a primary server might look like this:

failover peer "foo" {
  primary;
  address anthrax.rc.vix.com;
  port 519;
  peer address trantor.rc.vix.com;
  peer port 520;
  max-response-delay 60;
  max-unacked-updates 10;
  mclt 3600;
  split 128;
  load balance max seconds 3;
}

include "/etc/dhcpd.master";

The statements in the peer declaration are as follows:

Client classing

Clients can be separated into classes, and treated differently depending on what class they are in. This separation can be done either with a conditional statement, or with a match statement within the class declaration. It's possible to specify a limit on the total number of clients within a particular class or subclass that may hold leases at one time, and it's possible to specify automatic subclassing based on the contents of the client packet.

To add clients to classes based on conditional evaluation, you can specify a matching expression in the class statement:

class "ras-clients" {
 match if substring (option dhcp-client-identifier, 1, 3) = "RAS";
}
Note: Whether you use matching expressions or add statements (or both) to classify clients, you must always write a class declaration for any class that you use. If there will be no match statement and no in-scope statements for a class, the declaration should look like this:
class "ras-clients" {
}

Subclasses

In addition to classes, it's possible to declare subclasses. A subclass is a class with the same name as a regular class, but with a specific submatch expression which is hashed for quick matching. This is essentially a speed hack; the main difference between five classes with match expressions and one class with five subclasses is that it will be quicker to find the subclasses. Subclasses work as follows:

class "allocation-class-1" {
  match pick-first-value (option dhcp-client-identifier, hardware);
}

class "allocation-class-2" {
  match pick-first-value (option dhcp-client-identifier, hardware);
}

subclass "allocation-class-1" 1:8:0:2b:4c:39:ad;
subclass "allocation-class-2" 1:8:0:2b:a9:cc:e3;
subclass "allocation-class-1" 1:0:0:c4:aa:29:44;

subnet 10.0.0.0 netmask 255.255.255.0 {
  pool {
    allow members of "allocation-class-1";
    range 10.0.0.11 10.0.0.50;
  }
  pool {
    allow members of "allocation-class-2";
    range 10.0.0.51 10.0.0.100;
  }
}

The data following the class name in the subclass declaration is a constant value to use in matching the match expression for the class. When class matching is done, the server will evaluate the match expression and then look the result up in the hash table. If it finds a match, the client is considered a member of both the class and the subclass.

Subclasses can be declared with or without scope. In the above example, the sole purpose of the subclass is to allow some clients access to one address pool, while other clients are given access to the other pool, so these subclasses are declared without scopes. If part of the purpose of the subclass were to define different parameter values for some clients, you might want to declare some subclasses with scopes.

In the above example, if you had a single client that needed some configuration parameters, while most didn't, you might write the following subclass declaration for that client:

subclass "allocation-class-2" 1:08:00:2b:a1:11:31 {
  option root-path "samsara:/var/diskless/alphapc";
  filename "/tftpboot/netbsd.alphapc-diskless";
}

In this example, we've used subclassing as a way to control address allocation on a per-client basis. However, it's also possible to use subclassing in ways that aren't specific to clients — for example, to use the value of the vendor-class-identifier option to determine what values to send in the vendor-encapsulated-options option. An example of this is shown Vendor-encapsulated options in the DHCP options entry.

Per-class limits on dynamic address allocation

You may specify a limit to the number of clients in a class that can be assigned leases. The effect of this will be to make it difficult for a new client in a class to get an address. Once a class with such a limit has reached its limit, the only way a new client in that class can get a lease is for an existing client to relinquish its lease, either by letting it expire, or by sending a DHCPRELEASE packet. Classes with lease limits are specified as follows:

class "limited-1" {
  lease limit 4;
}

This will produce a class in which a maximum of four members may hold a lease at one time.

Spawning classes

It's possible to declare a spawning class, a class that automatically produces subclasses based on what the client sends. The reason that spawning classes were created was to make it possible to create lease-limited classes on the fly. The envisioned application is a cable-modem environment where the ISP wishes to provide clients at a particular site with more than one IP address, but doesn't wish to provide such clients with their own subnet, nor give them an unlimited number of IP addresses from the network segment to which they are connected.

Many cable modem head-end systems can be configured to add a Relay Agent Information option to DHCP packets when relaying them to the DHCP server. These systems typically add a circuit ID or remote ID option that uniquely identifies the customer site. To take advantage of this, you can write a class declaration as follows:

class "customer" {
  spawn with option agent.circuit-id;
  lease limit 4;
}

Now whenever a request comes in from a customer site, the circuit ID option will be checked against the class's hash table. If a subclass is found that matches the circuit ID, the client will be classified in that subclass and treated accordingly. If no subclass is found matching the circuit ID, a new one will be created and logged in the dhcpd.leases file, and the client will be classified in this new class. Once the client has been classified, it will be treated according to the rules of the class, including, in this case, being subject to the per-site limit of four leases.

The use of the subclass spawning mechanism isn't restricted to relay agent options; this particular example is given only because it's a fairly straightforward one.

Combining match, match if, and spawn with

In some cases, it may be useful to use one expression to assign a client to a particular class, and a second expression to put it into a subclass of that class. This can be done by combining the match if and spawn with statements, or the match if and match statements. For example:

class "jr-cable-modems" {
  match if option dhcp-vendor-identifier = "jrcm";
  spawn with option agent.circuit-id;
  lease limit 4;
}

class "dv-dsl-modems" {
  match if option dhcp-vendor-identifier = "dvdsl";
  spawn with option agent.circuit-id;
  lease limit 16;
}

This lets you have two classes that both have the same spawn with expression without getting the clients in the two classes confused with each other.

Dynamic DNS updates

The DHCP server has the ability to dynamically update the Domain Name System. Within the configuration files, you can define how you want the Domain Name System to be updated. These updates are RFC 2136 compliant, so any DNS server supporting RFC 2136 should be able to accept updates from the DHCP server.

Two DNS update schemes are currently implemented, and another is planned. The two that are currently available are the ad-hoc DNS update mode and the interim DHCP-DNS interaction draft update mode. If and when the DHCP-DNS interaction draft and the DHCID draft make it through the IETF standards process, there will be a third mode, which will be the standard DNS update method. The DHCP server must be configured to use one of the two currently-supported methods, or not to do dns updates. This can be done with the ddns-update-style configuration parameter.

The ad-hoc Dynamic DNS update scheme

Note: The ad-hoc Dynamic DNS update scheme is now deprecated and doesn't work. In future releases of the ISC DHCP server, this scheme will not likely be available. The interim scheme works, allows for failover, and should now be used. The following description is left here for informational purposes only.

The ad-hoc Dynamic DNS update scheme implemented in this version of the ISC DHCP server is a prototype design, which doesn't have much to do with the standard update method that is being standardized in the IETF DHC working group, but rather implements some very basic, yet useful, update capabilities. This mode doesn't work with the failover protocol because it doesn't account for the possibility of two different DHCP servers updating the same set of DNS records.

For the ad-hoc DNS update method, the client's FQDN is derived in two parts. First, the hostname is determined. Then, the domain name is determined, and appended to the hostname.

The DHCP server determines the client's hostname by first looking for a ddns-hostname configuration option, and using that if it's present. If no such option is present, the server looks for a valid hostname in the FQDN option sent by the client. If one is found, it is used; otherwise, if the client sent a host-name option, that is used. Otherwise, if there's a host declaration that applies to the client, the name from that declaration will be used. If none of these applies, the server will not have a hostname for the client, and will not be able to do a DNS update.

The domain name is determined from the ddns-domainname configuration option. The default configuration for this option is:

option server.ddns-domainname = config-option domain-name;

So if this configuration option isn't configured to a different value (overriding the above default), or if a domain-name option hasn't been configured for the client's scope, then the server will not attempt to perform a DNS update.

The client's fully qualified domain name, derived as we have described, is used as the name on which an “A” record will be stored. The A record will contain the IP address that the client was assigned in its lease. If there is already an A record with the same name in the DNS server, no update of either the A or PTR records will occur; this prevents a client from claiming that its hostname is the name of some network server. For example, if you have a fileserver called fs.sneedville.edu, and the client claims its hostname is fs, no DNS update will be done for that client, and an error message will be logged.

If the A record update succeeds, a PTR record update for the assigned IP address will be done, pointing to the A record. This update is unconditional; it will be done even if another PTR record of the same name exists. Since the IP address has been assigned to the DHCP server, this should be safe.

Note: The current implementation assumes clients only have a single network interface. A client with two network interfaces will see unpredictable behavior. This is considered a bug, and will be fixed in a later release. It may be helpful to enable the one-lease-per-client parameter so that roaming clients don't trigger this same behavior.

The DHCP protocol normally involves a four-packet exchange: first the client sends a DHCPDISCOVER message, then the server sends a DHCPOFFER, then the client sends a DHCPREQUEST, then the server sends a DHCPACK. In the current version of the server, the server will do a DNS update after it has received the DHCPREQUEST, and before it has sent the DHCPACK. It sends the DNS update only if it hasn't sent one for the client's address before, in order to minimize the impact on the DHCP server.

When the client's lease expires, the DHCP server (if it is operating at the time, or when next it operates) will remove the client's A and PTR records from the DNS database. If the client releases its lease by sending a DHCPRELEASE message, the server will likewise remove the A and PTR records.

The interim DNS update scheme

The interim DNS update scheme operates mostly according to several drafts that are being considered by the IETF and are expected to become standards, but aren't yet standards, and may not be standardized exactly as currently proposed. These are:

Because our implementation is slightly different than the standard, we will briefly document the operation of this update style here.

The first point to understand about this style of DNS update is that unlike the ad-hoc style, the DHCP server doesn't necessarily always update both the A and the PTR records. The FQDN option includes a flag which, when sent by the client, indicates that the client wishes to update its own A record. In that case, the server can be configured either to honor the client's intentions or ignore them. This is done with the statement allow client-updates or the statement ignore client-updates. By default, client updates are allowed.

If the server is configured to allow client updates, then if the client sends a fully-qualified domain name in the FQDN option, the server will use that name the client sent in the FQDN option to update the PTR record. For example, let us say that the client is a visitor from the radish.org domain, whose hostname is jschmoe. The server is for the example.org domain. The DHCP client indicates in the FQDN option that its FQDN is jschmoe.radish.org.. It also indicates that it wants to update its own A record. The DHCP server therefore doesn't attempt to set up an A record for the client, but does set up a PTR record for the IP address that it assigns the client, pointing at jschmoe.radish.org. Once the DHCP client has an IP address, it can update its own A record, assuming that the radish.org DNS server will allow it to do so.

If the server is configured not to allow client updates, or if the client doesn't want to do its own update, the server will simply choose a name for the client from either the fqdn option (if present) or the hostname option (if present). It will use its own domain name for the client, just as in the ad-hoc update scheme. It will then update both the A and PTR record, using the name that it chose for the client. If the client sends a fully-qualified domain name in the fqdn option, the server uses only the leftmost part of the domain name — in the example above, jschmoe instead of jschmoe.radish.org.

Further, if the ignore client-updates directive is used, then the server will in addition send a response in the DHCP packet, using the FQDN Option, that implies to the client that it should perform its own updates if it chooses to do so. With deny client-updates, a response is sent which indicates the client may not perform updates.

Also, if the use-host-decl-names configuration option is enabled, then the host declaration's hostname will be used in place of the hostname option, and the same rules will apply as described above.

The other difference between the ad-hoc scheme and the interim scheme is that with the interim scheme, a method is used that allows more than one DHCP server to update the DNS database without accidentally deleting A records that shouldn't be deleted nor failing to add A records that should be added. The scheme works as follows:

The interim DNS update scheme is called “interim” for two reasons. First, it doesn't quite follow the drafts. The current versions of the drafts call for a new DHCID RRtype, but this isn't yet available. The interim DNS update scheme uses a TXT record instead. Also, the existing ddns-resolution draft calls for the DHCP server to put a DHCID RR on the PTR record, but the interim update method doesn't do this. It is our position that this isn't useful, and we are working with the author in hopes of removing it from the next version of the draft, or better understanding why it is considered useful.

In addition to these differences, the server also doesn't update very aggressively. Because each DNS update involves a round trip to the DNS server, there's a cost associated with doing updates even if they don't actually modify the DNS database. So the DHCP server tracks whether or not it has updated the record in the past (this information is stored on the lease) and doesn't attempt to update records that it thinks it has already updated.

This can lead to cases where the DHCP server adds a record, and then the record is deleted through some other mechanism, but the server never again updates the DNS because it thinks the data is already there. In this case, the data can be removed from the lease through operator intervention, and once this has been done, the DNS will be updated the next time the client renews.

Dynamic DNS update security

When you set your DNS server up to allow updates from the DHCP server, you may be exposing it to unauthorized updates. To avoid this, you should use TSIG signatures, a method of cryptographically signing updates using a shared secret key. As long as you protect the secrecy of this key, your updates should also be secure. Note, however, that the DHCP protocol itself provides no security, and that clients can therefore provide information to the DHCP server which the DHCP server will then use in its updates, with the constraints described previously.

The DNS server must be configured to allow updates for any zone that the DHCP server will be updating. For example, let us say that clients in the sneedville.edu domain will be assigned addresses on the 10.10.17.0/24 subnet. In that case, you will need a key declaration for the TSIG key you will be using, and also two zone declarations: one for the zone containing A records that will be updates and one for the zone containing PTR records—for ISC BIND, something like this:

key DHCP_UPDATER {
  algorithm HMAC-MD5.SIG-ALG.REG.INT;
  secret pRP5FapFoJ95JEL06sv4PQ==;
};

zone "example.org" {
     type master;
     file "example.org.db";
     allow-update { key DHCP_UPDATER; };
};

zone "17.10.10.in-addr.arpa" {
     type master;
     file "10.10.17.db";
     allow-update { key DHCP_UPDATER; };
};

You will also have to configure your DHCP server to do updates to these zones. To do so, you need to add something like this to your dhcpd.conf file:

key DHCP_UPDATER {
  algorithm HMAC-MD5.SIG-ALG.REG.INT;
  secret pRP5FapFoJ95JEL06sv4PQ==;
};

zone EXAMPLE.ORG. {
  primary 127.0.0.1;
  key DHCP_UPDATER;
}

zone 17.127.10.in-addr.arpa. {
  primary 127.0.0.1;
  key DHCP_UPDATER;
}

The primary statement specifies the IP address of the name server whose zone information is to be updated.

Note: The zone declarations have to correspond to authority records in your name server; in the above example, there must be an SOA record for example.org. and for 17.10.10.in-addr.arpa.. For example, if there were a subdomain foo.example.org with no separate SOA, you couldn't write a zone declaration for foo.example.org. Also keep in mind that zone names in your DHCP configuration should end in a period (.); this is the preferred syntax. If you don't end your zone name in a period, the DHCP server will figure it out. Also note that in the DHCP configuration, zone names aren't encapsulated in quotes where there are in the DNS configuration.

You should choose your own secret key, of course. The ISC BIND 8 and 9 distributions come with a program for generating secret keys called dnssec-keygen. The version that comes with BIND 9 is likely to produce a substantially more random key, so we recommend you use that one even if you aren't using BIND 9 as your DNS server. If you are using BIND 9's dnssec-keygen, the above key would be created as follows:

dnssec-keygen -a HMAC-MD5 -b 128 -n USER DHCP_UPDATER

If you're using the BIND 8 dnskeygen program, the following command will generate a key as seen above:

dnskeygen -H 128 -u -c -n DHCP_UPDATER

You may wish to enable logging of DNS updates on your DNS server. To do so, you might write a logging statement like the following:

logging {
     channel update_debug {
          file "/var/log/update-debug.log";
          severity  debug 3;
          print-category yes;
          print-severity yes;
          print-time     yes;
     };
     channel security_info    {
          file "/var/log/named-auth.info";
          severity  info;
          print-category yes;
          print-severity yes;
          print-time     yes;
     };

     category update { update_debug; };
     category security { security_info; };
};

You must create the /var/log/named-auth.info and /var/log/update-debug.log files before starting the name server. For more information on configuring ISC BIND, consult the documentation that accompanies it.

Reference: events

There are three kinds of events that can happen regarding a lease, and it's possible to declare statements that occur when any of these events happen. These events are the commit event, when the server has made a commitment of a certain lease to a client, the release event, when the client has released the server from its commitment, and the expiry event, when the commitment expires.

To declare a set of statements to execute when an event happens, you must use the on statement, followed by the name of the event, followed by a series of statements to execute when the event happens, enclosed in braces. Events are used to implement DNS updates, so you should not define your own event handlers if you are using the built-in DNS update mechanism.

The built-in version of the DNS update mechanism is in a text string towards the top of server/dhcpd.c. If you want to use events for things other than DNS updates, and you also want DNS updates, you will have to start out by copying this code into your dhcpd.conf file and modifying it.

Reference: declarations

Reference: allow and deny

The allow and deny statements can be used to control the response of the DHCP server to various sorts of requests. The allow and deny keywords actually have different meanings depending on the context. In a pool context, these keywords can be used to set up access lists for address allocation pools. In other contexts, the keywords simply control general server behavior with respect to clients based on scope. In a non-pool context, the ignore keyword can be used in place of the deny keyword to prevent logging of denied requests.

Allow, deny, and ignore in scope

The following usages of allow and deny will work in any scope, although it isn't recommended that they be used in pool declarations.

Allow and deny within pool declarations

The uses of the allow and deny keywords shown in the previous section work pretty much the same way whether the client is sending a DHCPDISCOVER or a DHCPREQUEST message: an address will be allocated to the client (either the old address it's requesting, or a new address) and then that address will be tested to see if it's okay to let the client have it. If the client requested it, and it's not okay, the server will send a DHCPNAK message. Otherwise, the server will simply not respond to the client. If it is okay to give the address to the client, the server will send a DHCPACK message.

The primary motivation behind pool declarations is to have address allocation pools whose allocation policies are different. A client may be denied access to one pool, but allowed access to another pool on the same network segment. In order for this to work, access control has to be done during address allocation, not after address allocation is done.

When a DHCPREQUEST message is processed, address allocation simply consists of looking up the address the client is requesting and seeing if it's still available for the client. If it is, then the DHCP server checks both the address pool permit lists and the relevant in-scope allow and deny statements to see if it's okay to give the lease to the client. In the case of a DHCPDISCOVER message, the allocation process is done as described previously in the ADDRESS ALLOCATION section.

When you're declaring permit lists for address allocation pools, the following syntaxes are recognized following the allow or deny keywords:

known-clients;
If specified, this statement either allows or prevents allocation from this pool to any client that has a host declaration (i.e., is known). A client is known if it has a host declaration in any scope, not just the current scope.
unknown-clients;
If specified, this statement either allows or prevents allocation from this pool to any client that has no host declaration (i.e., isn't known).
members of "class";
If specified, this statement either allows or prevents allocation from this pool to any client that is a member of the named class.
dynamic bootp clients;
If specified, this statement either allows or prevents allocation from this pool to any bootp client.
authenticated clients;
If specified, this statement either allows or prevents allocation from this pool to any client that has been authenticated using the DHCP authentication protocol. This isn't yet supported.
unauthenticated clients;
If specified, this statement either allows or prevents allocation from this pool to any client that has not been authenticated using the DHCP authentication protocol. This isn't yet supported.
all clients;
If specified, this statement either allows or prevents allocation from this pool to all clients. This can be used when you want to write a pool declaration for some reason, but hold it in reserve, or when you want to renumber your network quickly, and thus want the server to force all clients that have been allocated addresses from this pool to obtain new addresses immediately when they next renew.
after time;
If specified, this statement either allows or prevents allocation from this pool after a given date. This can be used when you want to move clients from one pool to another. The server adjusts the regular lease time so that the latest expiry time is at the given time + min-lease-time. A short min-lease-time enforces a step change, whereas a longer min-lease-time allows for a gradual change. The time is either a number of seconds since the Unix epoch, a UTC time string (e.g. 4 2007/08/24 09:14:32), or a string with time zone offset in seconds (e.g., 4 2007/08/24 11:14:32 -7200).

Reference: parameters

Setting parameter values using expressions

Sometimes it's helpful to be able to set the value of a DHCP server parameter based on some value that the client has sent. To do this, you can use expression evaluation. For information on writing expressions, see the DHCP Conditional Evaluation entry. To assign the result of an evaluation to an option, define the option as follows:

my-parameter = expression ;

For example:

ddns-hostname = binary-to-ascii (16, 8, "-",
                                 substring (hardware, 1, 6));

Reserved leases

It's often useful to allocate a single address to a single client, in approximate perpetuity. Host statements with fixed-address clauses exist to a certain extent to serve this purpose, but because host statements are intended to approximate “static configuration”, they suffer from not being referenced in a litany of other Server Services, such as dynamic DNS, failover, “on events” and so forth.

If a standard dynamic lease, as from any range statement, is marked reserved, then the server will allocate this lease only to the client it is identified by (be that by client identifier or hardware address).

In practice, this means that the lease follows the normal state engine, enters ACTIVE state when the client is bound to it, expires, or is released, and any events or services that would normally be supplied during these events are processed normally, as with any other dynamic lease. The only difference is that failover servers treat reserved leases as special when they enter the FREE or BACKUP states; each server applies the lease into the state it may allocate from, and the leases aren't placed on the queue for allocation to other clients. Instead they may only be “found” by client identity. The result is that the lease is only offered to the returning client.

Care should probably be taken to ensure that the client has only one lease within a given subnet that it is identified by.

Leases may be set reserved either through OMAPI, or through the infinite-is-reserved configuration option (if this is applicable to your environment and mixture of clients).

It should also be noted that leases marked reserved are effectively treated the same as leases marked bootp.

See also:

Contributing author:

dhcpd.conf was written by Ted Lemon under a contract with Vixie Labs. Funding for this project was provided by Internet Systems Consortium. Information about Internet Systems Consortium can be found at http://www.isc.org.