IPv6 Migration Checklist for Enterprise Networks: Addressing to Security

Our engineering team walks through an ipv6 migration checklist for enterprise networks, from addressing plans and dual-stack rollout to DHCPv6 and security controls.

Quick answer: what belongs in an IPv6 migration checklist for enterprise networks?

A working ipv6 migration checklist for enterprise networks follows a clear order. Start with the addressing plan, enable dual-stack next, decide how clients obtain addresses through DHCPv6 or SLAAC, and only then tighten security. The governing rule is simple: existing IPv4 services must keep running throughout the project.

Projects that skip ahead tend to run into trouble. In several incidents our team investigated, the root cause was turning on IPv6 before anyone had written down an addressing plan. A network configured from memory breaks quickly.

The sections below follow that same order. Treat them as a working list. Tick off each step as your team completes it, and keep a copy in your change-management record so the next engineer can see what was done and why.

Inventory first, because no plan survives an unknown network

Begin with an accurate inventory of every device, its firmware version, and the applications it hosts. Add a column that records IPv6 support. A datasheet that says “IPv6 supported” tells you very little, since support often depends on the exact software release.

A surprising number of devices advertise the feature while lacking the function you actually need, such as IPv6 ACLs or DHCPv6 relay. Test those specific functions on a bench unit before trusting the marketing line.

Define scope early as well. Campus access, WAN links, data center fabric, and internet-facing services each behave differently. Printers and building cameras on an undocumented VLAN are easy to overlook. A checklist built from the real device list, not from memory, is the most reliable way to close that kind of gap.

The addressing plan comes before everything else

The addressing plan is the map that governs how IPv6 space is divided. Engineers who grew up with 10.x and 192.168.x ranges often find the mental shift difficult. IPv6 space is enormous, so the goal is orderly design rather than conserving every address.

The process usually begins with a prefix, obtained from your ISP or, if you hold your own allocation, from your Regional Internet Registry. From there you divide the prefix by site, by VLAN, and by service class. The widely adopted convention is a /64 for every LAN subnet, as described in RFC 4291. Mechanisms such as SLAAC depend on that /64 boundary, so deviating from it creates problems that are painful to unwind later.

Document your conventions as well. Decide which address the gateway uses, which block the servers occupy, and where the DHCPv6 pool sits. When something fails at two in the morning, a consistent convention saves hours of troubleshooting.

Here is a minimal gateway configuration on a Cisco IOS router for one VLAN:

interface Vlan10
 ipv6 address 2001:db8:10:10::1/64
 ipv6 enable

The 2001:db8::/32 range is reserved for documentation, so replace it with your real prefix. The ipv6 enable command guarantees a link-local address even before a global address is assigned, and both routing and neighbor discovery depend on that link-local address.

Dual-stack, and why rollout order matters

Dual-stack means devices run IPv4 and IPv6 side by side. Most enterprises choose this path, and it is also where faults are most often blamed on the wrong layer. Roll out from the inside outward: enable IPv6 on the core and distribution layers first, then extend it to access.

The logic is straightforward. Once the core can route IPv6, the layers beneath it have a valid path to follow. If you enable access first, clients receive IPv6 addresses with nowhere to send traffic. Users then experience intermittent connectivity that is genuinely hard to diagnose, because the symptoms look random.

Routing needs equal attention. Verify IPv6 routing with OSPFv3 or BGP for IPv6, and monitor both protocol families together. A route missing on one side produces one-directional failures, which are difficult to spot in a packet capture.

DHCPv6 or SLAAC: who should assign addresses to clients?

This is the question we hear most often, and the right answer depends on how much control you need. SLAAC, defined in RFC 4862, lets each device build its own address from advertised prefixes. It requires no server, which is appealing, but it gives you little visibility into which user holds which address.

Stateful DHCPv6, specified in RFC 8415, reverses that trade-off. You decide which address each client receives and you keep a log for later investigation. In audit-heavy environments it is usually the safer choice.

Many office networks combine the two. SLAAC handles addressing while DHCPv6 distributes DNS and other options. That keeps SLAAC’s simplicity while preserving a centralized configuration channel. If your organization faces strict internal audits, lean toward stateful DHCPv6 and accept the extra server to maintain.

Three deployment approaches and what each costs you

Not every network needs pure dual-stack. Dual-stack is the default for most enterprises, but tunneling and translation still appear in real projects, and each carries its own burden.

Tunneling mechanisms such as 6to4 and ISATAP are quick to stand up. Our team treats them as a temporary bridge only. When something fails, tracing packets through encapsulation layers consumes a great deal of time, and the failure modes are hard to predict.

NAT64 combined with DNS64 suits IPv6-only clients that still need IPv4-only resources. It works well for specific segments, but it adds the translation gateway as a dependency you must monitor, patch, and size correctly. Treat it as infrastructure, not as a quick fix.

Security controls that IPv4 habits tend to miss

This is where many operations teams relax too early. A common assumption is that IPv6 is just another address family, so existing rules can simply be copied across. In practice that assumption causes trouble.

Router Advertisement Guard blocks rogue devices from sending router advertisements, much as DHCP snooping blocks rogue DHCP servers on IPv4. Enable it on access ports. DHCPv6 snooping plays a comparable role for DHCPv6 traffic and should be deployed alongside Router Advertisement Guard wherever DHCPv6 is in use.

The most common self-inflicted outage involves ICMPv6 filtering. Teams sometimes block all ICMPv6 out of habit. That is a serious error, because ICMPv6 carries neighbor discovery and Path MTU Discovery. Blocking it indiscriminately produces failures that look exactly like random application timeouts. RFC 4890 lists the ICMPv6 message types that should remain permitted, and it is worth reading before you write any filter.

Firewalls need their own IPv6 policy rather than an inherited copy of the IPv4 rules. Syntax varies by vendor and software release, so verify against the manufacturer’s current documentation before you deploy anything to production.

Pilot on one VLAN before you scale

Resist the urge to expand before you have tested. Choose a single pilot VLAN, enable IPv6 there, and check reachability with ping6 and traceroute6. Confirm that AAAA records resolve to the correct addresses, because a misconfigured record can send users to a dead endpoint without any obvious error message.

At the same time, review internal applications for hard-coded IPv4 addresses. This is a recurring problem. Legacy software often has a fixed IP written into its configuration. After dual-stack goes live, such an application may keep reaching for the IPv4 address while the IPv6 path fails in a way that looks unrelated to the application itself.

Expand only after the pilot has run cleanly through at least one full working cycle, including a month-end batch if your business has one. That single test catches more problems than most lab exercises.

Mistakes we see again and again

Beyond the ICMPv6 problem covered above, three other errors recur across projects. The first is forgetting to update firewalls, ACLs, and monitoring for IPv6, which leaves a large share of traffic invisible to the team. Dashboards that show only IPv4 flows give a false sense of security.

The second is choosing a prefix that is too small because it seemed sufficient at the time. Renumbering later is expensive and disruptive. The third is skipping training, so nobody on the operations team knows how to read IPv6 logs when an incident finally occurs.

Plan and train first, then account for the cost

Migration cost depends heavily on how many devices need upgrades and how large the network is, so no single figure applies to every organization. Much of the expense falls on engineering and operations staff time rather than on hardware.

Investing in training and planning before rollout almost always costs less than repairing a poorly sequenced deployment afterward. Budget for the hours your engineers will spend learning, testing, and documenting, not just for the equipment line item.

Combining the migration with a Zero Trust approach

As IPv6 expands the number of globally reachable addresses, trusting a device simply because it sits inside the network becomes harder to defend. Every global address is a potential entry point, so segmentation based on location alone becomes fragile. This is a good moment to consider identity-based access control.

For background on moving from network-zone thinking to identity-centric thinking, review the security architecture resources at abctech.vn before you redraw your trust boundaries. If you want to compare current access control approaches, our colleagues have written about modern access control strategies, which can help you decide which model fits your environment.

Infrastructure and security teams also benefit from reviewing broader engineering resources together. Our partners at itacenter.vn and softsupplier.com cover adjacent operational topics, such as tooling for the monitoring stack, that often come up during a migration.

Frequently asked questions about IPv6 migration

Do we need to replace all our hardware for IPv6?

Usually not. Many recent devices support IPv6 in software, so a firmware upgrade is often enough. Genuinely old equipment typically needs replacing, and the inventory you built at the start of the project will show which devices fall into that category.

Should we turn off IPv4 once IPv6 is running?

For most enterprises we advise against switching IPv4 off in a hurry. Dual-stack keeps legacy services reachable. Consider removing IPv4 only after every application has been verified on IPv6 and you have a clear rollback plan that you have actually rehearsed.

How can we confirm a device fully supports IPv6?

Do not rely on marketing material. Check the vendor’s official documentation for the exact software version you run, then validate on a pilot VLAN. Basic IPv6 support and support for the security features you need are often two very different things.

Should an office network use SLAAC or DHCPv6?

If you need auditing and centralized control, choose stateful DHCPv6. If simplicity and fewer servers matter more, SLAAC paired with DHCPv6 for DNS is a balanced option. Let your organization’s real audit requirements make the final call, and document the decision so the reasoning survives staff changes.

Related topics