4.8. Network Renumbering With the mechanisms given by ICMPv6, renumbering a network in an IPv6 world may become a lot easier in the future. Currently there is not much operational experience. Renumbering a network means replacing an old prefix with a new prefix. This can become necessary for a number of reasons, a common one being a change of provider, which usually implies a change of prefix. Renumbering a network may encompass the following steps: Each link in the network must be assigned a subprefix from the new prefix before beginning the procedure. This is important for the overall process, in order to ensure proper configuration of all relevant devices and services such as routers, switches, interfaces, DNS, DHCP, etc. The DNS database must be updated with the addresses for interfaces from the new prefix, and addresses for interfaces from the old prefix must be removed. Obviously the changes to DNS must be coordinated with the changes to addresses assigned to interfaces. The propagation of this new information can be controlled by parameters such as the "Time to Live" (TTL) for DNS records and the update interval between primary and secondary DNS servers. Switches and routers are prepared for the new prefix. All necessary changes in the routing infrastructure for the new prefix are added in parallel to the old prefix, while the old prefix is still used for datagram services. This includes not only routers and switches, but also firewalls, ingress and egress filters, and all other filtering functions. For propagating subnet prefix information to routers, the IPv6 Prefix option for DHCPv6 (RFC 3633) may be used. In the case where hosts use Stateless autoconfiguration, the routers are not configured to advertise the prefix for autoconfiguration yet (meaning that the Autonomous Address Configuration flag is not set). This will be done once stable routing for the new prefix has been verified. All access lists, route maps, and other network configuration options (e.g., name services other than DNS) that use IP addresses should be checked to ensure that hosts and services that use the new prefix will behave as they did with the old one. Test and verify network infrastructure and routing for the new prefix. Advertise the new prefix outside of the corporate network. Configure all border defense systems accordingly to protect the new prefix from outside attacks. Assign addresses from the new prefix to interfaces on hosts while still retaining the addresses from the old prefix. If Stateless autoconfiguration is used, the "autonomous address-config" flag is set for the new prefix, so hosts configure addresses for the new prefix in addition to the old addresses. DHCP now assigns addresses from both prefixes if it is used. The new information can be propagated by using the DHCP Reconfigure message, which will cause every DHCP client to contact the DHCP server. The addresses from the new prefix will not be used until they are inserted into DNS. When the new prefix has been fully integrated into the network infrastructure and tested for stable operation, hosts, switches, and routers can begin using the new prefix. Once the transition has completed, the old prefix will not be in use in the network and can be removed step by step from DNS.
Special attention has to be given to applications and devices that do not get their IP addresses from DHCP or DNS, or that cache or store IP address information locally. This is a high-level view of a renumbering process and obviouslyas all network administrators know wellthere are many details and possible pitfalls to be considered. Thorough and careful planning of this process is a must. RFC 4192 describes this process. |