Network repro matrix script for comparing intermittent curl timing stalls across Macs

network_repro_matrix.sh

#!/usr/bin/env bash

set -euo pipefail

REQUESTS="${REQUESTS:-15}"
STALL_SECS="${STALL_SECS:-0.5}"
MAX_TIME="${MAX_TIME:-45}"
CONNECT_TIMEOUT="${CONNECT_TIMEOUT:-40}"
OUTFILE="${OUTFILE:-network_repro_$(date +%Y%m%d_%H%M%S).csv}"

usage() {
  cat <<'EOF'
Usage:
  ./network_repro_matrix.sh [requests]

Environment overrides:
  REQUESTS=20          Number of requests per target/mode
  STALL_SECS=0.5       Threshold used to mark a request as a stall
  MAX_TIME=45          curl --max-time
  CONNECT_TIMEOUT=40   curl --connect-timeout
  OUTFILE=results.csv  CSV output path

What it does:
  - tests several different sites
  - runs each site in normal mode and IPv4-only mode (-4)
  - records DNS, connect, TLS, first-byte, and total time
  - prints a short summary and writes raw rows to CSV
EOF
}

if [[ "${1:-}" == "-h" || "${1:-}" == "--help" ]]; then
  usage
  exit 0
fi

if [[ $# -ge 1 ]]; then
  REQUESTS="$1"
fi

if ! command -v curl >/dev/null 2>&1; then
  echo "curl is required" >&2
  exit 1
fi

if ! command -v python3 >/dev/null 2>&1; then
  echo "python3 is required" >&2
  exit 1
fi

targets=(
  "nytimes|https://static01.nyt.com/images/2026/04/15/multimedia/15DC-SURVEILLANCE-vkgp/15DC-SURVEILLANCE-vkgp-threeByTwoMediumAt2X.jpg?format=pjpg&quality=75&auto=webp&disable=upscale"
  "google|https://www.google.com/generate_204"
  "cloudflare|https://www.cloudflare.com/cdn-cgi/trace"
  "apple|https://www.apple.com/library/test/success.html"
  "example|https://example.com/"
)

tmpfile="$(mktemp)"
trap 'rm -f "$tmpfile"' EXIT

printf 'label,mode,run,http_code,remote_ip,exit_code,dns,connect,tls,starttransfer,total,stall\n' > "$OUTFILE"

echo "Writing raw results to $OUTFILE"
echo "Requests per target/mode: $REQUESTS"
echo "Stall threshold: ${STALL_SECS}s"
echo

for entry in "${targets[@]}"; do
  label="${entry%%|*}"
  url="${entry#*|}"

  for mode in default ipv4; do
    curl_args=()
    if [[ "$mode" == "ipv4" ]]; then
      curl_args+=("-4")
    fi

    echo "== $label [$mode] =="

    for run in $(seq 1 "$REQUESTS"); do
      fmt='http_code=%{http_code} remote_ip=%{remote_ip} dns=%{time_namelookup} connect=%{time_connect} tls=%{time_appconnect} start=%{time_starttransfer} total=%{time_total}'
      set +e
      result="$(
        curl -sS -o /dev/null \
          --max-time "$MAX_TIME" \
          --connect-timeout "$CONNECT_TIMEOUT" \
          "${curl_args[@]}" \
          -w "$fmt" \
          "$url" 2>&1
      )"
      curl_exit=$?
      set -e

      python3 - "$label" "$mode" "$run" "$curl_exit" "$STALL_SECS" "$OUTFILE" "$result" <<'PY'
import csv
import re
import sys

label, mode, run, curl_exit, stall_secs, outfile, result = sys.argv[1:]

fields = {
    "http_code": "000",
    "remote_ip": "",
    "dns": "",
    "connect": "",
    "tls": "",
    "start": "",
    "total": "",
}

for key in list(fields):
    m = re.search(rf"{key}=([^ \n]+)", result)
    if m:
        fields[key] = m.group(1)

try:
    total = float(fields["total"])
except Exception:
    total = float("inf")

stall = "yes" if total > float(stall_secs) else "no"

with open(outfile, "a", newline="") as f:
    writer = csv.writer(f)
    writer.writerow([
        label,
        mode,
        run,
        fields["http_code"],
        fields["remote_ip"],
        curl_exit,
        fields["dns"],
        fields["connect"],
        fields["tls"],
        fields["start"],
        fields["total"],
        stall,
    ])

summary = (
    f"run{run} total={fields['total']}s "
    f"dns={fields['dns']} connect={fields['connect']} "
    f"tls={fields['tls']} start={fields['start']} "
    f"ip={fields['remote_ip'] or '-'} code={fields['http_code']} "
    f"exit={curl_exit}"
)
if stall == "yes":
    summary += " STALL"

print(summary)
if curl_exit != "0":
    print(result.strip())
PY
    done

    echo
  done
done

python3 - "$OUTFILE" <<'PY'
import csv
import statistics
import sys
from collections import defaultdict

rows = list(csv.DictReader(open(sys.argv[1], newline="")))
groups = defaultdict(list)
for row in rows:
    groups[(row["label"], row["mode"])].append(row)

print("Summary:")
for key in sorted(groups):
    bucket = groups[key]
    totals = []
    stalls = 0
    connect_over_1 = 0
    start_over_1 = 0
    for row in bucket:
        try:
            total = float(row["total"])
            totals.append(total)
            if total > 0.5:
                stalls += 1
        except Exception:
            pass
        try:
            if float(row["connect"]) > 1.0:
                connect_over_1 += 1
        except Exception:
            pass
        try:
            if float(row["starttransfer"]) > 1.0:
                start_over_1 += 1
        except Exception:
            pass
    med = statistics.median(totals) if totals else float("nan")
    worst = max(totals) if totals else float("nan")
    print(
        f"- {key[0]:11s} {key[1]:7s} "
        f"stalls={stalls}/{len(bucket)} "
        f"median={med:.3f}s worst={worst:.3f}s "
        f"connect>1s={connect_over_1} start>1s={start_over_1}"
    )
PY

echo
echo "Done. Raw rows are in $OUTFILE"