| 2 Jan 2025 |
 hexa (clat on linux when) | enable debug logging for networkd and grep for dhcp to find out why it enables it | 14:54:18 |
 antifuchs | just noticed that I had vestigial dhcp relaying from that workaround activated | 14:54:50 |
| antifuchs | are you using useNetworkd or enabling manual networkd configuration? | 14:55:05 |
| antifuchs | * just noticed that I had vestigial dhcp relaying from that workaround activated; rebuilding | 14:55:30 |
 K900 | No useNetworkd here | 14:59:59 |
| hexa (clat on linux when) | systemd.network.enable | 14:58:33 |
| antifuchs | ic. ok, so disabling DHCP serving for systemd on all interfaces (even those without kea) does allow kea to listen on :67 | 14:59:59 |
| antifuchs | smells like an annoying trap tbh | 15:00:10 |
| hexa (clat on linux when) | well, given that the socket is bound to a wildcard address, that is to be expected | 15:00:26 |
| antifuchs | I mean, systemd manages to bind multiple times to the wildcard there too, so I assumed that would work | 15:02:12 |
| hexa (clat on linux when) | probably use SO_REUSEPORT | 15:03:26 |
| hexa (clat on linux when) | * probably uses SO_REUSEPORT | 15:03:43 |
| antifuchs | * ic. ok, so disabling DHCP serving for systemd-networkd on all interfaces (even those without kea) does allow kea to listen on :67 | 15:04:29 |
 Jassuko | If only systemd-networkd finally supported network namespaces as a first class citizen... -_- | 15:05:09 |
| hexa (clat on linux when) | ideally, yes | 15:08:20 |
| hexa (clat on linux when) | but make sure you can't live with a vrf instead | 15:08:28 |
| hexa (clat on linux when) | someone probably has to lobby yuwata a bit ð | 15:09:11 |
| hexa (clat on linux when) | https://github.com/systemd/systemd/pull/14915 | 15:09:13 |
| hexa (clat on linux when) | * someone probably has to lobby yuwata a bit to make netns support happen ð | 15:09:23 |
 adamcstephens | RIP | 15:28:36 |
| adamcstephens | * RIP that pR | 15:28:48 |
| adamcstephens | * RIP that PR | 15:28:50 |
| hexa (clat on linux when) | uh, yeah | 15:31:44 |
| hexa (clat on linux when) | * https://github.com/systemd/systemd/pull/14915 ðŠĶ | 15:31:53 |
| hexa (clat on linux when) | updated ð | 15:31:56 |
 kdn | can bonding 2x2.5GbE interface reduce the throughput? I'm observing ~400 mbps from desktop/laptop > switch1 > switch2 > BOND > router, but not desktop <> switch1 > laptop | 17:12:18 |
| K900 | Very possibly | 17:12:47 |
| kdn | it's 802.3ad bonding | 17:12:53 |
| K900 | You'd have to measure every hop separately to know | 17:12:57 |
| K900 | Especially cheap hardware I'd expect to not handle this correctly | 17:13:06 |
