relay-debug.sh

#!/usr/bin/env bash
# Debug relay transfer performance across different relay servers.
#
# Runs a provide+fetch pair for each relay, with full trace logging.
# Results go into ./logs/relay-debug-<timestamp>/.

set -euo pipefail

RELAYS=(
    "euc1-1.relay.n0.iroh-canary.iroh.link"
    "use1-1.relay.n0.iroh-canary.iroh.link"
)
SIZE="${SIZE:-1m}"
MODE="${MODE:-bidi}"
if [ -n "${TRANSFER:-}" ]; then
    :
else
    target_dir=$(cargo metadata --format-version 1 --no-deps 2>/dev/null | jq -r '.target_directory')
    candidate="${target_dir}/release/examples/transfer"
    if [ -x "$candidate" ]; then
        TRANSFER="$candidate"
    else
        echo "ERROR: release transfer example not found at ${candidate}"
        echo "Build it first: cargo build --release --example transfer --all-features"
        echo "or set path to transfer binary with TRANSFER=/path/to/transfer"
        exit 1
    fi
fi

ts=$(date +%y%m%d-%H%M%S)
OUTDIR="./logs/relay-debug-${ts}"
mkdir -p "$OUTDIR"

export RUST_LOG="${RUST_LOG:-iroh=trace,noq=trace}"

for relay_host in "${RELAYS[@]}"; do
    relay_url="https://${relay_host}"
    label="${relay_host%%.*}"
    echo "=== ${label} (${relay_url}) ==="

    ping -c 2 -W 2 "$relay_host" 2>&1 | tee "${OUTDIR}/${label}-ping.txt" || true
    echo ""

    # Start provide in background
    "$TRANSFER" provide \
        --relay-only \
        --relay-url "$relay_url" \
        --output json \
        2>"${OUTDIR}/${label}-provide.log" \
        >"${OUTDIR}/${label}-provide.json" &
    provide_pid=$!

    # Wait for the endpoint id to appear in the JSON output
    endpoint_id=""
    for i in $(seq 1 30); do
        sleep 0.5
        if [ -f "${OUTDIR}/${label}-provide.json" ]; then
            endpoint_id=$(grep -o '"endpoint_id":"[^"]*"' "${OUTDIR}/${label}-provide.json" 2>/dev/null | head -1 | cut -d'"' -f4 || true)
            if [ -n "$endpoint_id" ]; then
                break
            fi
        fi
    done

    if [ -z "$endpoint_id" ]; then
        echo "ERROR: could not get endpoint_id from provide after 15s"
        kill "$provide_pid" 2>/dev/null || true
        wait "$provide_pid" 2>/dev/null || true
        continue
    fi

    echo "endpoint_id: ${endpoint_id}"

    if ! kill -0 "$provide_pid" 2>/dev/null; then
        echo "ERROR: provide died before fetch could start. Check ${OUTDIR}/${label}-provide.log"
        wait "$provide_pid" 2>/dev/null || true
        continue
    fi

    echo "fetching..."

    # Run fetch. Trace logs go to file, text output to terminal + file.
    "$TRANSFER" fetch "$endpoint_id" \
        --relay-only \
        --relay-url "$relay_url" \
        --remote-relay-url "$relay_url" \
        ${SIZE:+--size "$SIZE"} \
        --mode "$MODE" \
        --output text \
        2>"${OUTDIR}/${label}-fetch.log" |
        tee "${OUTDIR}/${label}-fetch.txt" ||
        echo "fetch exited with $?"

    # Stop provide
    kill "$provide_pid" 2>/dev/null || true
    wait "$provide_pid" 2>/dev/null || true

    echo ""
done

echo "--- Summary ---"
for relay_host in "${RELAYS[@]}"; do
    label="${relay_host%%.*}"
    echo ""
    echo "=== ${label} ==="
    grep "MiB in\|KiB in\|GiB in" "${OUTDIR}/${label}-fetch.txt" 2>/dev/null || echo "(no results)"
    grep "time=" "${OUTDIR}/${label}-ping.txt" 2>/dev/null | tail -1 || true
done

echo ""
echo "Logs in: ${OUTDIR}"