test

System Administration Guide: IP Services

How DHCP Works

The DHCP server must first be installed and configured by a system administrator. During configuration, the administrator enters information about the network that clients will need to operate on the network. After this information is in place, clients are able to request and receive network information.

The sequence of events for DHCP service is shown in the following diagram. The numbers in circles correlate to the numbered items in the description following the diagram.

Figure 7–1 Sequence of Events for DHCP Service

Diagram shows communication sequence between a DHCP client and server. The legend that follows the diagram explains the sequence.

LEGEND:

  1. The client discovers a DHCP server by broadcasting a discover message to the limited broadcast address (255.255.255.255) on the local subnet. If a router is present and configured to behave as a BOOTP relay agent, the request is passed to other DHCP servers on different subnets. The client's broadcast includes its unique ID, which in the Solaris DHCP implementation, is derived from the client's Media Access Control (MAC) address. On an Ethernet network, the MAC address is the same as the Ethernet address.

    DHCP servers that receive the discover message can determine the client's network by looking at the following information:

    • Which network interface did the request come in on? This tells the server that the client is either on the network to which the interface is connected, or that the client is using a BOOTP relay agent connected to that network.

    • Does the request include the IP address of a BOOTP relay agent? When a request passes through a relay agent, the relay agent inserts its address in the request header. When the server detects a relay agent address, it knows that the network portion of the address indicates the client's network address because the relay agent must be connected to the client's network.

    • Is the client's network subnetted? The server consults the netmasks table to find the subnet mask used on the network indicated by the relay agent's address or the address of the network interface that received the request. Once the server knows the subnet mask used, it can determine which portion of the network address is the host portion, and then select an IP address appropriate for the client. (See netmasks(4) for information on netmasks.)

  2. After they determine the client's network, DHCP servers select an appropriate IP address and verify that the address is not already in use. The DHCP servers then respond to the client by broadcasting an offer message that includes the selected IP address and information about services that can be configured for the client. Each server temporarily reserves the offered IP address until it can determine if the client will use it.

  3. The client selects the best offer (based on the number and type of services offered) and broadcasts a request that specifies the IP address of the server that made the best offer. The broadcast ensures that all the responding DHCP servers know the client has chosen a server, and those servers not chosen can cancel the reservations for the IP addresses they had offered.

  4. The selected server allocates the IP address for the client, stores the information in the DHCP data store, and sends an acknowledgement (ACK) to the client. The acknowledgement message contains the network configuration parameters for the client. The client uses ping to test the IP address to make sure no other system is using it, then continues booting to join the network.

  5. The client monitors the lease time, and when a set period of time has elapsed, the client sends a new message to the chosen server to increase its lease time.

  6. The DHCP server that receives the request extends the lease time if it still adheres to the local lease policy set by the administrator. If the server does not respond within 20 seconds, the client broadcasts a request so that one of the other DHCP servers can extend the lease.

  7. When the client no longer needs the IP address, it notifies the server that it is releasing the IP address. This can happen during an orderly shutdown and can also be done manually.