I'm using OpenWrt as dumb/bridged AP and the wiki encourages disabling the DHCP in favor of the upstream router, and that is entirely reasonable. However, in case of upstream DHCP server's outage for whatever reason, I still want the clients to get IP from the OpenWrt device similar to how most OEM firmware behaves if upstream connection is down/missing.
As such, I've jerry-rigged the behavior on my own. I'm posting this to ask if there's a better way and if my method could lead to some unforeseen pitfalls.
For my method, I use the default behavior of force under config dhcp being disabled to my advantage. This being the default has cause me problems every now and then, but in this case OpenWrt can work as a "hot backup" if upstream connection has been disrupted. I keep the default lan interface unchanged (network address is different from OpenWrt defaults) and add a new interface with the same br-lan device with static IP of the LAN to keep it accessible. In summary, it has two IPs for lan interface, one with default static IP and DHCP enabled, the other with upstream network-specific static IP; in effect, clients get IP from upstream router, but if that connection is gone, OWrt serves IP so the AP's LuCi/ssh is easily accessible in both cases.
There are two concerns I have with this setup:
- Upstream router doesn't do IPv6, so the OpenWrt DHCPv6 is taking up the slack. If upstream did have v6, will DHCPv6 be idling similar to v4?
- What if I have multiple OpenWrt AP with this kind of "dynamic DHCP" set up? Should I be concerned of multiple DHCP servers wrecking havoc if upstream is gone? OEM dynamic DHCP thingies has already caused me such pain with their APs advertising DHCP even when they're connected over WAN port. I hope similar crapstorm does not result from my own hands.