I have tried two setups. Unfortunately in setup a) proxmox is dropping all UDP coming from SBC with a length greater of about 1500 bytes (do not know exactly). Meaning all shorter packages do pass, larger packages are dropped as described below. Setup b) works perfectly, I have 0% drops, no matter the length. I do not have enough hardware ports on the proxmox server to do it without VLANs. If this is not my mistake, I think this bug is very serious.
Using tcpdump I tracked the first network controller that did show the shorter packages but NOT the longer ones. They are lost between tap100i6 and vmbr0 on proxmox. They are "bridged" or connected (listed in one line) in the PFSense VM Setup. So it seems to be an error completely related to proxmox and not the pfsense VM.
PS: I am usin virtio everywhere.
PS2: I was checking the package flow on every network adapter using "tcpdump -nqt -s 0 -A -i [adapter] port 5060 and host [expected IP]".
PS3: PFSense was complaining "UDP, bad length 1697 > 1472" or similar. However, the internet says that this message was legit. When checking on tap100i6, tcpdump did not complain about bad length and was mentioning the same length of 1697.
a) Cisco Session Border Controller (Swisscom SBC Software, 192.168.99.17) => USW-EnterpriseXG-24 (VLAN Tag 10) => Intel E810-XXVDA2 (enp129s0f0) => Proxmox (vmbr0, enp129s0f0 is member, VLAN Aware, VLAN ID 10) => VM IN (PFsense, net3, VLAN Tag 10) => VM Out (PFsense, net6) => Proxmox Virtual Adapter (tap100i6) => Proxmox (vmbr0, enp129s0f0 is member, VLAN Aware, VLAN ID 6) => USW-EnterpriseXG-24 (VLAN Tag 6) => Centro Business Modem => Swisscom SBC (Sip provider)
b) Cisco Session Border Controller (Swisscom SBC Software, 192.168.99.17) => USW-EnterpriseXG-24 (VLAN Tag 10) => Intel E810-XXVDA2 (enp129s0f0) => Proxmox (vmbr0, enp129s0f0 is member, VLAN Aware, VLAN ID 10) => VM IN (PFsense, net3, VLAN Tag 10) => VM Out (PFsense, net6) => Proxmox (vmbr3, No VLAN, onboard eno2np1 is member) => Centro Business Modem => Swisscom SBC (Sip provider)
Using tcpdump I tracked the first network controller that did show the shorter packages but NOT the longer ones. They are lost between tap100i6 and vmbr0 on proxmox. They are "bridged" or connected (listed in one line) in the PFSense VM Setup. So it seems to be an error completely related to proxmox and not the pfsense VM.
PS: I am usin virtio everywhere.
PS2: I was checking the package flow on every network adapter using "tcpdump -nqt -s 0 -A -i [adapter] port 5060 and host [expected IP]".
PS3: PFSense was complaining "UDP, bad length 1697 > 1472" or similar. However, the internet says that this message was legit. When checking on tap100i6, tcpdump did not complain about bad length and was mentioning the same length of 1697.
a) Cisco Session Border Controller (Swisscom SBC Software, 192.168.99.17) => USW-EnterpriseXG-24 (VLAN Tag 10) => Intel E810-XXVDA2 (enp129s0f0) => Proxmox (vmbr0, enp129s0f0 is member, VLAN Aware, VLAN ID 10) => VM IN (PFsense, net3, VLAN Tag 10) => VM Out (PFsense, net6) => Proxmox Virtual Adapter (tap100i6) => Proxmox (vmbr0, enp129s0f0 is member, VLAN Aware, VLAN ID 6) => USW-EnterpriseXG-24 (VLAN Tag 6) => Centro Business Modem => Swisscom SBC (Sip provider)
b) Cisco Session Border Controller (Swisscom SBC Software, 192.168.99.17) => USW-EnterpriseXG-24 (VLAN Tag 10) => Intel E810-XXVDA2 (enp129s0f0) => Proxmox (vmbr0, enp129s0f0 is member, VLAN Aware, VLAN ID 10) => VM IN (PFsense, net3, VLAN Tag 10) => VM Out (PFsense, net6) => Proxmox (vmbr3, No VLAN, onboard eno2np1 is member) => Centro Business Modem => Swisscom SBC (Sip provider)