Kea Flow Diagrams

These flow diagrams describe Kea's DHCPv4 server implementation, and they may be useful for system administrators. To design a configuration that results in clients getting the intended addresses and options, it is important to understand the sequence of request-processing steps. For example, Kea iterates looking for a suitable address, and conditionally accepts the first available address, so the order in which addresses are evaluated matters.

It is also useful to understand Kea's processing logic because there are configuration choices which can make the process far more efficient. Kea is very flexible, so it can be applied to very different use cases and in different environments. In an environment where throughput and efficiency are a priority, the administrator can choose to limit some of the processing steps. For example, it is possible to limit the number of different client identifiers Kea evaluates in looking for a host reservation, or even to skip the step of checking for host reservations.

These diagrams are focused on those aspects of Kea processing that will be most useful to operators. The diagrams illustrate DHCPv4 request processing, but most of the logic applies equally to DHCPv6. Following the title of each diagram is a Kea version number. Kea behavior has evolved over time, and the diagrams document the behavior as of the Kea version indicated. These diagrams are provided in the Kea source tree in UML (source), PNG, and SVG formats.

Main Loop

The main loop is common to both DHCPv4 and DHCPv6 servers.

_images/main-loop.svg

The DHCP server main loop

DHCPv4 Packet Processing

DHCPv4 packet processing evaluates the type DHCP message: Discover, Request, Release, Decline, or Inform. This diagram shows the general, high-level flow for processing an inbound client DHCP packet from receipt to the server's response.

_images/packet4.svg

DHCPv4 packet processing

DHCPREQUEST Processing

The following diagrams focus on DHCPREQUEST processing. This chart gives an overview of the process, from subnet selection to checking for host reservations to evaluating client classes. Finally, before acknowledging the lease, the options are evaluated and added to the message.

_images/request4.svg

DHCPREQUEST processing

DHCPv4 Subnet Selection

Subnet selection is the process of choosing a subnet that is topologically appropriate for the client. When the selected subnet is a member of a shared network, the whole shared network is selected. During subnet selection the client class may be checked more than once while iterating through subnets, to determine whether it is permitted in the selected subnet.

_images/select4.svg

DHCPv4 subnet selection

DHCPv4 Special Case of Double-Booting

After subnet selection and before lease allocation, the DHCPv4 server handles the special case of clients restarting with an image provided by PXE boot or bootp. The Lease Request box is expanded below.

_images/assign-lease4.svg

DHCPv4 lease assignment

DHCPv4 Lease Allocation

The first diagram below illustrates the details of processing the client request, showing the renewal of an existing lease, the assignment of a reserved lease, and the allocation of an unreserved lease.

_images/request4-lease.svg

DHCPREQUEST lease allocation

The second diagram shows the algorithm used to validate a requested lease or select a new address to offer. The right-hand side of the diagram shows how a new address is selected, when a new lease is required and the client has neither a requested address nor a reservation. When a new lease is required and Kea iterates over pools and subnets, it starts with the subnet selected above in the subnet selection process.

_images/requestLease4.svg

The requestLease4 algorithm

Note

Declined addresses are included in the statistic for assigned addresses, so the \(assigned + free = total\) equation is true.

Lease States

This diagram illustrates the different lease states, including the free one, where no lease object exists.

_images/lease-states.svg

Lease states

Checking for Host Reservations

The allocation engine checks for host reservations after selecting a subnet. The following diagram shows the details of that operation. Subnet selection is based on network topology. Host reservations are primarily for assigning options, and options are evaluated after subnet selection. However, if client classes are added in the host reservation, those are also evaluated against the selected subnet in a further check. Kea includes several options to skip checking for host reservations, which can make this process much more efficient if reservations are not being used.

Note

To find a free lease, the allocation engine begins by evaluating the most recently used subnet. The current subnet depends on the history of prior queries.

_images/currentHost4.svg

Host reservation evaluation

Building the Options List

Before sending a response, options are added:
  • evaluate required client classes

  • build the configured option list

  • append requested options

  • append requested vendor options

  • append basic options

_images/buildCfgOptionList.svg

The buildCfgOptionList (build configured option list) algorithm

_images/appendRequestedOptions.svg

The appendRequestedOptions (append requested options) algorithm

_images/appendRequestedVendorOptions.svg

The appendRequestedVendorOptions (append vendor requested options) algorithm

How Kea Recognizes the Same Client In Different DHCP Messages

_images/recognizing-same-client.svg

RADIUS workflows for lease allocation

_images/radius.svg