iperf-grpc-tcpdump

name	iperf-grpc-tcpdump
description	Run synchronized iperf3 tests from the isolated userspace client while capturing server-side tcpdump over the gRPC capture service and simultaneous LAN/WAN tcpdumps on the active userspace firewall. Use this when debugging throughput collapse, TCP resets, or tuple corruption on the userspace dataplane.

iperf3 + gRPC tcpdump

Use this skill when the task is to reproduce or debug userspace dataplane forwarding failures with synchronized captures.

This skill targets the isolated loss userspace lab:

• client: loss:cluster-userspace-host
• firewalls: loss:xpf-userspace-fw0, loss:xpf-userspace-fw1
• server capture gRPC endpoint: 172.16.80.200:50051

The helper script:

• detects the active firewall automatically
• starts server-side capture through grpcurl
• starts LAN and WAN tcpdump on the active firewall
• runs iperf3 from cluster-userspace-host
• saves artifacts under /tmp

Run:

.codex/skills/iperf-grpc-tcpdump/scripts/capture_iperf.sh --family 4
.codex/skills/iperf-grpc-tcpdump/scripts/capture_iperf.sh --family 6

Useful options:

.codex/skills/iperf-grpc-tcpdump/scripts/capture_iperf.sh --family 4 --parallel 4 --duration 8
.codex/skills/iperf-grpc-tcpdump/scripts/capture_iperf.sh --family 6 --parallel 1 --duration 5 --cport 55340

Artifacts:

• server-grpc.txt
• fw-lan.txt
• fw-wan.txt
• fw-stats-before.txt
• fw-stats-after.txt
• iperf.json
• summary.txt

Workflow:

1. Run the helper script for the family and stream count you want.
2. Read summary.txt first.
3. If throughput collapses after a strong first interval:
   • compare fw-wan.txt and server-grpc.txt
   • look for port drift, duplicate tuples, RSTs, or missing replies
4. Use fw-stats-before.txt and fw-stats-after.txt to confirm whether the active firewall actually traversed the userspace dataplane.

Assumptions:

• grpcurl, incus, and iperf3 are installed locally
• the isolated userspace cluster already exists
• the server-side gRPC capture service is reachable on 172.16.80.200:50051

Do not use this skill to change cluster config or rebuild VMs. It is for repeatable capture and analysis only.