Upgraded to 1.5-202406290019 - DHCP server not serving leases? (failed to open socket - interface is not running)

Bugs
1

I just upgraded a host running an older VyOS rolling build (1.5-rolling-202309170024) to the current VyOS rolling build (1.5-rolling-202406290019).

However, on reboot - it no longer seemed to be handing out DHCP leases to guests correctly.

On the serial console, I did notice these messages - not sure if that’s actually, or safe to ignore (i.e. red herring):

error: no suitable video mode found.
error: no suitable video mode found.
Booting in blind modeBooting in blind mode

[    0.182474] DMAR: [Firmware Bug]: No firmware reserved region can cover this RMRR [0x000000003e2e0000-0x000000003e2fffff], contact BIOS vendor for fixes
[    0.396634] fail to initialize ptp_kvm
[   21.636776] iTCO_wdt iTCO_wdt: can't request region for resource [mem size 0x00000004 64bit]
[   22.266357] EDAC pnd2: Failed to register device with error -19.

Here’s some of the logfiles, as well as the VyOS configuration:

  • dmesg output from the box (gist)
  • DHCP server logs (gist)
  • VyOS configuration (gist)

In the DHCP server logs, I do see a bunch of messages about failed to open socket, and the interface <name> is not running?

Does anybody know what might be happening, or if this is a recent breakage, or something that’s changed in configuration?

2

It would really help to provide some basic config output.
What does your DHCP server config look like before (202309170024), and after (202406290019)?

If you run tcpdump do you see any DHCP packets? Is the DHCP daemon running (ps aux) etc.

Newer version of 1.5 are using the Kea DHCP Server instead of the old ISC one (it’s end of life) so potentially you’ve hit a bug in the upgrade path.

3

OK, I deleted the broken version (202406290019), and booted into the older working version (202309170024).

I then downloaded the latest rolling image again (202406300022), and went through the upgrade process again.

I booted into it again - and somehow, DHCP did seem to work (although I had to wait for some devices to renew, I guess).

For reference - here are the two configurations:

Is it possible that the upgrade with that sightly older rolling (only one day - at 202406290019) was broken somehow?

Or do you spot some lurking issues in the newer upgraded config?

4

The 20240630 version have “no changes” compared to 20240629 according to:

But the changes is only regarding VyOS internal changes - could have been some Debian changes (deb-packages) in between which unfortunately is not visible at all through the changelogs or release tracking.

If possible could you paste “dpkg -l” from both the 20240629 and 20240630 install - that should tell if any of the Debian packages have been changed in between and if so which ones?