Unusual Unifi DHCP Server Issues

1

I have a Unifi deployment with a UDM-Pro with 5 VLANs. The DHCP server (hosted on the UDM Pro) has progressively stopped working on all but 1 VLAN. The setup has been working fine for well over a year and since it only started to be apparent when new devices joined, I don’t know exactly at what point it started failing.

  • Devices that have had an existing IP address and maintained connectivity still work fine.
  • New devices joining the network fail and get a self assigned IP.
  • Manually configuring IP addresses on the clients works fine.
  • Seems to be impacting Unifi device adoption now, as rebooting a switch which was adopted before, now results in a failed adoption.

I have checked the following, but I cannot seem to get the DHCP to give IP addresses.

  1. Confirmed via packet capture that clients are sending DHCP requests to the gateway.
listening on wlp192s0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
22:14:19.960781 IP 0.0.0.0.bootpc > 255.255.255.255.bootps: BOOTP/DHCP, Request from 8e:46:3d:0f:d4:da, length 286
22:14:22.053263 IP 0.0.0.0.bootpc > 255.255.255.255.bootps: BOOTP/DHCP, Request from 8e:46:3d:0f:d4:da, length 286

  1. Confirmed via packet capture on the UDM-Pro that the requests are being received, but the gateway is not replying.
    listening on br0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
    15:23:48.023411 IP 0.0.0.0.68 > 255.255.255.255.67: BOOTP/DHCP, Request from fa:2d:6f:fd:5b:ab, length 300
    15:23:56.207157 IP 0.0.0.0.68 > 255.255.255.255.67: BOOTP/DHCP, Request from fa:2d:6f:fd:5b:ab, length 300

  2. Confirmed the DHCP server is running on the UDM Pro, but getting “No Address Available” errors.
    2026-03-20T18:44:49-04:00 UDM-Pro dnsmasq-dhcp[1817854]: DHCPDISCOVER(br0) ae:db:93:61:af:5c no address available

  3. Confirmed addresses should be available, at least according to the Unifi UI.

    image
    image1675×518 90.6 KB

  4. Confirmed the switch ports are allowing all VLANs.
    Also confirmed that the ports on the netgear switch have VLAN 1 as the untagged and all other VLANs tagged (for the ports going to the Unifi Switch and AP at least). I am not using any port profiles.

    Example:

    image
    image369×437 18.1 KB

  5. Device firmware is up to date

    image
    image686×254 18.9 KB

  6. DHCP lease time is set to 1 Day (86400s).

    Topology
    UDM Pro is connected to an Aggregation switch (STP0) which has a Pro Max Switch (STP4096) and a Netgear switch (STP4096) attached. There is one more Unifi switch connected to to the Netgear switch with an STP of 8195. I specify the STP, as this is the last thing I remember playing with before I noticed problems. But I don’t see any STP blocking issues, and the settings seem valid based on Unifi’s documentation.

    image
    image784×903 39.4 KB

Any further ideas to help debug ?

2

Do you have any firewall rules stopping the devices from talking to the gateway? If you bypass the Netgear doe the devices get an IP?

3

Appreciate your response @LTS_Tom! I don’t have any firewall rules within a VLAN, and I can see the DHCP request hit the gateway in the packet captures. One of my access points is plugged in upstream of the Netgear so that also doesn’t seem to be the issue.

Error seems to be the “No Address Available” from the DHCP logs. The only weird thing I noticed is that if I list out the ARP table on the UDM-Pro, I see about 220 entries which would indicate that the subnet space is filling up (maybe from clients with MAC randomization), but in the UI it seems to lots of space. Is it possible that it’s not clearing out the leases correctly? Would there be any relation between the ARP table to the DHCP leases ? Not sure if DNSMasq is using that somehow to track the space?

4

Yes, there is typically a 1:1 relationship between ARP table and DHCP leases. While they are separate and live on completely different caches a device has to receive an IP address via a lease before it appears in the ARP table.

My concern is ARP table filling up which seems to match the “no address available” from your UDM Pro. Switch adoption failing seems likely because it’s not receiving an IP address.

Further more what seems even stranger this is manifesting on all your VLANs. I could understand a chatty IoT device with MAC randomization exhausting the available pool but that is for 1 VLAN.

The fact it happens on all seems to point to either stale leases that are never released or released too late back into the pool, or some other issue with the DHCP server itself. One place I’d look is the duration of the lease, maybe lower it more aggressively to say 6 hours and see what happens. Also if you can at the same time clear the table of old leases that are not currently in use.

5

I was online with UniFi Support this morning and after switching the default subnet to a /23, it started handing out addresses normally. Therefore there is indeed a synchronization issue with the DHCP server and the addresses are not being released correctly. It’s also not properly reflected in the Network UI which shows plenty of addresses available. They are going to look into it more and get back to me. I will post anything I hear in case others have this problem as well.

2 Likes