A client has a remote office wherein an L3 switch is configured with ip helper-addresses (on multiple vlan interfaces) configured to obtain their IP address from the head office. But due to the often-times unreliable WAN link, end users are sometimes unable to get an IP address.
My solution was to configure the remote L3 switch (3750) as the backup DHCP server and to enable it when the WAN link goes down. I thought that it was a simple transition of by just removing/reinstating the IP helper-address commands, I'll have control on where users can get their IP addresses (local switch or remote DHCP).
However, I was surprised to find out that even with an ip helper-address defined, the users were getting their IP addresses from the switch. I was under the impression that once an ip helper-address is defined, it will take precedence over the built-in IOS DHCP server.
Is this the true behavior of the ip helper-address feature or am I hitting a bug?
Switch IOS version is: 12.2(25r)SEE3
In support of the answer from Glen: yes this is the proper and expected behavior and is NOT a bug. The helper address does not pre-emptivly take over from the locally configured DHCP service. If the local switch has DHCP configured and also has the helper address configured then the switch will forward the DHCP request to the remote DHCP server and will also generate a response to the client to offer its own address pool. You have not correctly understood how it works and Glen has correctly explained it.
All the ip helper address does is take a dhcp broadcast request and make it a unicast to the defined address . The clients will usually take the first device which answers first which will be the directly attached switch because it is closer and will respond faster . You would have to turn off the service on the switch first before the primary would take over.