DNS Resolver - Unbound - Not able to resolve DHCP leases until it is saved

[slackadelic]

So, this odd issue popped up where I was not able to resolve DHCP leases within the DNS Resolver.

What I found was an article speaking about how Unbound is not able to resolve new leases as it doesn't know about them until you 'save' the configuration again.

As a test, I logged into OPNsense, went to the DNS Resolver page, and submitted a save, even though no changes on my side had happened.

Bingo.. once it applied, active DHCP clients were now resolvable on my network.

[fichtner]

@StopTrain reported that not setting up IPv6 unbreaks unbound

[fichtner]

@StopTrain hey Matt, close this or investigate further?

[8191]

Investigating a different issue with this feature I can now tell what's the reason for this one, I'd say:

The DHCP leases are mapped as static host overrides within unbound. The "magic" is done within etc/inc/unbound.inc:560-575, which (afaik) only get's called when applying changes in the DNS Resolver.
@fichtner Is there a mechanism which calls the config scripts regularly?

[fichtner]

There is quite a bit of complication going on. There is a daemon called dhcpleases, which knows the PID of the DNS daemon so the config can be reloaded using SIGHUP, but I don't know how the host file is regenerated other than probably doing it in the daemon itself, not sure if some code is missing since the GUI code was forked 9 months after the dhcpleases code disappeared...

https://github.com/opnsense/core/blob/master/src/etc/inc/system.inc#L465
https://github.com/opnsense/ports/blob/master/opnsense/dhcpleases/files/dhcpleases.c#L491

I've ran a test with Matt using the newer daemon's code but it was unsuccessful.

[fichtner]

Matt, @AdSchellevis has revorked this for the development version of 16.1.5. I've tested this, looked pretty good now. Can you retest when it's out?

[slackadelic]

I most certainly can.

[fichtner]

This is in since a while, I'd like to vote for close and worst case opening a new bug. The behaviour of the resolver seems to be better now and original behaviour may not apply. :)

[fichtner]

timeout