Today in "how did my life decisions lead me here?", diagnosing hanging TCP connections from a NATed container through another layer of NAT on my router up to the internet. With two different kernels and three network stacks involved (calico, the bare linux stack, and FreeBSD/OPNSense).
Of course the issue is pretty random (though it seems to always end up happening), and I can only really narrow it down to an IP range where I have some MBps of sustained traffic all through the chain, making packet captures... On the large side.
This all feels pretty manual. Any tool recommendations to
- dump conntrack states in an exploitable format?
- match tcpdump traces with recorded conntrack states?
- align and compare tcpdump traces at multiple points in a chain?
- analyze issues in tcp sessions that I've captured?
So far I've used #conntrack #wireshark #tcpdump, but still haven't managed to nail it.
#Linux #Opnsense #sysadmin