Captive Portal system hang

[tawmu]

Hi there,

Has anyone else seen an issue with Captive Portal in OPNsense causing high CPU usage?

We run LAN gaming events and two events in a row now  (July/October) we've seen OPNsense begin to hang and drop packets once our captive portal entries hit somewhere around 650 - 700 user entries. The issue is resolved when we are eventually able to login to the Web UI and disable the captive portal. We're currently running release 24.7 - hardware is an Intel E-2314, 32GB RAM, and an Intel X520-DA2 card.

Symptoms are 100% system CPU usage and 5-15% packet loss, seemingly on processes kernel{if_io_tqq_0}, kernel{if_io_tqq_1}, kernel{if_io_tqq_2}, and kernel{if_io_tqq_3}. Roughly 650-700 captive portal entries and 2.5Gbps of traffic flowing between SFP+ interfaces (hoping to up it to 10Gbps for our next event...).

The Captive Portal on OPNsense simply serves a HTML redirect page - we handle authentication in an external app that makes an API call to OPNsense to add the user's session in.

Apologies if there's any mistakes on the process names or for lack of other information - it's quite difficult to troubleshoot when you've got 1,000 gamers experiencing packet loss at a live event :)