You dump scopes for Lattice. You expect the failover relationship to follow.
It does not.
Microsoft DHCP failover is a partner relationship — load-balance or hot-standby, hashed pair or split scope. Lattice does not import partners. Live leases do not import either. What comes over is the plan you map: prefixes, subnets, pools, reservations (Replace Windows DHCP, Migrate).
What “failover” meant on Windows
Two servers. A named relationship. Scopes shared under Microsoft’s rules. Discover on A and Ack on B only works when that partnership is healthy. Operators learn this during DC upgrades, broken replication, and PowerShell jobs that deactivate a scope and push that deactivation to the partner.
That pain is real. It is also not Lattice’s HA model.
When the partnership is sick, MMC lies in a special way. One server thinks it owns free addresses. The other thinks the same. Clients get duplicate Offers or long silence. Your runbook grows a “check failover health” section that has nothing to do with prefixes.
What Lattice HA is
Same pool on every node. Identical dhcpd processes share the lease store. Discover on one node and Ack on another is the same lease — by design (HA, Replace Windows DHCP).
Paid HA shape: three Ubuntu VMs, one Postgres, one VIP. No partner server. No split scope.
Do not read Lattice as cheaper Microsoft failover. Do not read it as an Infoblox Grid.
Community is one node. Same binaries. A Starter license adds multi-node dhcpd on that shared database and three-node VIP when you need it (HA).
Helpers point at the VIP (or a single Community node). Trusted relays list those helper source IPs. There is no second Microsoft server to keep “in sync” after the cut.
Load-balance vs hot-standby vs same pool
Microsoft load-balance splits free addresses by hash. Hot-standby keeps a partner quiet until failure. Both are partner models. Both assume two Windows DHCP servers that know each other.
Lattice same-pool HA assumes one catalog and identical dhcpd processes. There is no free-address hash to import. There is no “relationship name” field in the migrate API.
If your runbook says “recreate failover after import,” delete that line for Lattice. Replace it with: verify VIP, verify trusted relays, verify one helper, prove a lease, then remove Windows from the helpers list.
What to bring on cutover
- Dump scopes and reservations (PowerShell mapping is on Migrate).
- Map to prefix → subnet → pool → reservation.
- POST
/api/v1. - List trusted relays. Point helpers at Lattice (or the VIP).
- Turn off Windows DHCP on that VLAN — do not dual-helper.
Printers that must keep an IP become reservations. Failover partners stay behind. Live leases stay behind. Occupancy starts at the first Ack Lattice grants (Migrate, Who held this IP — occupancy).
If you are mid-DC migration with DHCP still on domain controllers, park Windows DHCP first, then cut to Lattice when ready — see Move Windows DHCP off your DCs before the OS upgrade rather than mixing both projects in one weekend.
Dual-helper is still wrong
During cutover, two DHCP authorities on one VLAN make duplicate Offers. That was true on Windows. It stays true with Lattice. Park, cut, validate — do not “temporary dual-helper for safety.”
Operators reach for dual-helper because partner failover taught them redundancy equals two answers. Lattice redundancy is two (or three) nodes behind one VIP answering from one catalog. The relay still sees one destination.
After the first Ack
Helpdesk Search finds the lease Lattice granted — not the lease Windows held five minutes ago.
Occupancy history begins at cutover. Do not expect “who had .55 last month” from Lattice until Lattice has been granting.
That is the honest trade: you leave partner state and live lease history on Windows. You gain one catalog, same pool on every node, and a cutover you can explain in five bullets.
What helpdesk hears after cutover
“Failover is down” is not a Lattice ticket. There is no partner relationship to repair. Ask: Is the VIP up? Are trusted relays listed? Is Windows still on a helper?
Those three questions replace a Microsoft failover health dashboard for this product.
Soft next step
Finish the dump of scopes and reservations while the Windows pair is still healthy (How to dump Windows DHCP scopes). Read Replace Windows DHCP for the one-catalog compare. Join Community for 1,000 managed IPs, or start the 30-day Starter trial in the portal when you want same-pool HA instead of a partner relationship (Pricing, Signup).
latticeddi.com