One detail usually explains "works at first, fails at renewal". A client's first lease uses broadcasts that the router relays (IP helper) to the DHCP server. A renewal at 50% of the lease is a unicast REQUEST sent straight from the client to the server. If a firewall, ACL or Dynamic ARP Inspection/DHCP snooping on the path allows the relayed traffic but not direct UDP 67/68 between the client subnet and the server, renewals fail until the client falls back to rebinding.
To prove where it breaks, capture both ends at the same moment. This is Microsoft's own DHCP troubleshooting method:
- Install Wireshark on one affected client and on the DHCP server. Start both captures at the same time.
- On the client run
ipconfig /renew(or wait for a natural renewal). Filter the client capture withdhcp, and note the Transaction ID. - On the server capture, filter with
dhcp.id == <that Transaction ID>. - If the REQUEST leaves the client but never reaches the server, the network path is dropping it. Check firewall/ACL rules, DHCP snooping and DAI.
- If the server receives it but sends no ACK, or sends a NAK, check the server's events under Applications and Services Logs > Microsoft > Windows > DHCP-Server and the logs in
%windir%\System32\Dhcp. - If the ACK reaches the client but it still drops off, look at the client's System log and NIC power settings.
- If you use DHCP Failover or split scopes, check that both servers agree on the lease, since inconsistent lease activity often shows up there.
Microsoft's TSS tool also collects all of this in one go: TSS.ps1 -Scenario NET_DHCPcli on the client and NET_DHCPsrv on the server.
Are the affected clients on a different subnet from the DHCP server (using a relay)?
If the capture pinpoints the drop, please consider accepting it as the answer. If both captures look clean, let me know and I'll be happy to keep helping.
Reference: