hub2 firmware inbound 80/443 problems

[#1]

Hub Two firmware 0.11.01.11071-PN, updated 5-Jun-2026. Intermittent inbound TCP 80/443 after approximately 1–2 minutes idle. First SYN is delayed/dropped for approximately 10 seconds; once delivered, the LAN server completes the TCP/HTTP transaction in ~20 ms. Subsequent connections are immediate until another idle period.

Connections within the LAN are at us / ms latency.

I have 1Gbps fibre to the premises. Tried rebooting the router many times. Any ideas?

[#2]

@dcldesign 

First SYN is delayed/dropped for approximately 10 seconds; once delivered, the LAN server completes the TCP/HTTP transaction in ~20 ms

What is the NIC the server is using and platform you are using ? It's as if it (NIC) has to come out of standby, is there a power save mode active.

What if you keep a repeating ping on the server IP Address, does that help?

Edit: another thought try this https://www.thinkbroadband.com/broadband-quality-monitoring 

[#3]

Hi thanks for replying. It's definitely not power-saving on the NIC. The setup is router --> nginx reverse proxy --> server endpoints. Many ports are forwarded as necessary for the servers (80/443/imap/smtp/DNS/etc, etc). There are roughly 4 requests per second incoming as a baseline traffic + bursts as they occur. Everything is debian of some flavour or another (roughly 20 VMs + other devices on the network). At most times not much traffic flows in/out to WAN.

Until the last few days the setup has been fine, but it has occasionally thrown fits like this before for 24-48 hours. I have tried all manner of rebooting, including shutting literally everything down and waiting 15 minutes, then starting up the router, then the hosts, then the VMs etc. It seems that 24 hours or so after a problem like this occurs, it seems to fix itself. I half expect to find some weird lease or cache retention on the router, but nothing has ever come to light. I've also extensively tested with tools like nc, tcpdump, dig, ping, etc. The results are always inconclusive. I have tried forwarding port 80 to a single IP and bypassing the nginx etc, and it makes literally no difference. 

I'd like to know why this happens! Or, shall I just buy another NIC and do the routing myself with openwrt?

[#4]

How is you LAN setup, I assume everything can not be connected to the Hub Two, so switches must be involved, are they daisy chained or star connected to the Hub which would be best? Any routers back to back creating double NAT.

Are there any other devices which could be using some of the upload bandwidth, cloud save, one drive?

[#5]

There is one managed switch (it's configuration is as a simple passthrough device, it's not filtering or anything) and no other router, no double NAT or anything like that. Everything is on a single logical /24 segment. No other devices are using upload bandwidth except in the single or double digit Kibps range. I guess what I'm saying is that there's no real load on the router or even the network itself. I can saturate the link with a NAS transfer for example, all packets ping between devices with very low latency. I just don't think the problem is on the LAN.