Hey Neutron Team. All good with this, we were chasing a red herring 😊 Undocumented network change… --Karl. From: Karl Kloppenborg <kkloppenborg@resetdata.com.au> Date: Saturday, 27 April 2024 at 7:14 pm To: openstack-discuss@lists.openstack.org <openstack-discuss@lists.openstack.org> Subject: [Neutron] Openvswitch forcing MTU 1450 Hey Neutron team. I have an issue where within the VXLAN network of tenancies the expected 1500 MTU is honoured but despite all MTU in the underlay being 9k and all networks and paths set to 1500, Openvswitch caps the MTU transiting north/south to 1450. This is causing considerable pain to some of our customers and despite trying to force the openvswitche bridges request_mtu to 1500, and all interfaces look right, it’s still caping at 1450. Note, TCP dumps show the pings set to 1500 getting to the network namespace on the openvswitch agent, so it appears something between the nat/snat process. VM1 private -> VM2 private (mtu 1500) = works VM1 private -> VM2 public (mtu 1500) = fails VM -> External (mtu 1500) = fails External Host -> VM (mtu 1500) = fails Adusting to 1422 on ping sizes works. Anyone experienced this? I cannot seem to make heads or tails of it. Thanks, Karl.