fix(tooling): harden internode transport benchmark setup (#3037)

* refactor(config): centralize internode transport constants

* fix(bench): guard all ripgrep calls behind dry-run check

Move require_cmd rg and metrics collection inside the non-dry-run
path so that --dry-run works on hosts without rg installed.

* feat(tooling): cross-platform protoc setup for Linux and macOS

Make install-protoc.sh support Linux (x86_64, aarch64) alongside
macOS, and bump CI protoc from 29.3 to 33.1 to match the version
required by the gproto build script.

* fix(bench): record internode baseline error counts

* fix(skill): correct YAML frontmatter formatting for release-version-bump

* chore(ci): bump protoc version to 34.1

* fix(tooling): bump protoc 33.1 to 34.1 in install script, restore SKILL.md description

---------

Co-authored-by: Henry Guo <marshawcoco@users.noreply.github.com>

scripts/install-protoc.sh

@@ -1,22 +1,42 @@
 #!/usr/bin/env bash
-# Install protoc 33.1 on macOS
+# Install protoc 34.1 on macOS and Linux
 
 set -e
 
-PROTOC_VERSION="33.1"
+PROTOC_VERSION="34.1"
 ARCH=$(uname -m)
 INSTALL_DIR="${HOME}/.local/bin"
 PROTOC_BIN="${INSTALL_DIR}/protoc"
 
-# Select download URL based on architecture
-if [ "$ARCH" = "arm64" ]; then
-    # Apple Silicon (M1/M2/M3)
-    PROTOC_URL="https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-osx-aarch_64.zip"
-elif [ "$ARCH" = "x86_64" ]; then
-    # Intel Mac
-    PROTOC_URL="https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-osx-x86_64.zip"
+# Detect OS
+OS="$(uname -s)"
+
+# Select download URL based on OS and architecture
+if [ "$OS" = "Darwin" ]; then
+    if [ "$ARCH" = "arm64" ]; then
+        # Apple Silicon (M1/M2/M3)
+        PROTOC_URL="https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-osx-aarch_64.zip"
+    elif [ "$ARCH" = "x86_64" ]; then
+        # Intel Mac
+        PROTOC_URL="https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-osx-x86_64.zip"
+    else
+        echo "Error: Unsupported macOS architecture $ARCH"
+        exit 1
+    fi
+elif [ "$OS" = "Linux" ]; then
+    if [ "$ARCH" = "x86_64" ]; then
+        PROTOC_URL="https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-linux-x86_64.zip"
+    elif [ "$ARCH" = "aarch64" ]; then
+        PROTOC_URL="https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-linux-aarch_64.zip"
+    else
+        echo "Error: Unsupported Linux architecture $ARCH"
+        exit 1
+    fi
 else
-    echo "Error: Unsupported architecture $ARCH"
+    echo "Error: Unsupported OS $OS"
+    echo "On Windows, install protoc via:"
+    echo "  choco install protoc --version=${PROTOC_VERSION}"
+    echo "  or download from https://github.com/protocolbuffers/protobuf/releases"
     exit 1
 fi
 

scripts/run_internode_transport_baseline.sh

@@ -125,7 +125,7 @@ setup_output() {
     OUT_DIR="target/bench/internode-transport-$(date +%Y%m%d-%H%M%S)"
   fi
   mkdir -p "${OUT_DIR}"
-  echo "scenario,endpoint,workload,concurrency,size,status,throughput,requests_per_sec,avg_latency,log_file,run_dir" > "${OUT_DIR}/summary.csv"
+  echo "scenario,endpoint,workload,concurrency,size,status,throughput,requests_per_sec,avg_latency,error_count,log_file,run_dir" > "${OUT_DIR}/summary.csv"
   if [[ -n "${INTERNODE_METRICS_URL}" ]]; then
     echo "scenario,workload,concurrency,size,metric,operation,before,after,delta" > "${OUT_DIR}/internode_metric_deltas.csv"
   fi
@@ -253,7 +253,7 @@ append_object_summary() {
   awk -F',' -v scenario="${scenario}" -v endpoint="${endpoint}" -v workload="${workload}" -v run_dir="${run_dir}" '
     NR == 1 { next }
     {
-      printf "%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s\n", scenario, endpoint, workload, $3, $1, $4, $5, $6, $7, $8, run_dir
+      printf "%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s\n", scenario, endpoint, workload, $3, $1, $4, $5, $6, $7, $8, $9, run_dir
     }
   ' "${src}" >> "${OUT_DIR}/summary.csv"
 }

scripts/run_object_batch_bench.sh

@@ -187,7 +187,7 @@ setup_output() {
   fi
   mkdir -p "$OUT_DIR"
   SUMMARY_CSV="$OUT_DIR/summary.csv"
-  echo "size,tool,concurrency,status,throughput,requests_per_sec,avg_latency,log_file" > "$SUMMARY_CSV"
+  echo "size,tool,concurrency,status,throughput,requests_per_sec,avg_latency,error_count,log_file" > "$SUMMARY_CSV"
 }
 
 resolve_bucket() {
@@ -214,6 +214,27 @@ collect_metrics() {
   echo "${throughput:-N/A},${reqps:-N/A},${latency:-N/A}"
 }
 
+collect_error_count() {
+  local log_file="$1"
+  local count
+
+  count="$(
+    { rg -io '(errors?|failures?)[[:space:]:=]+[0-9]+' "$log_file" || true; } \
+      | awk '{ value = $0; gsub(/[^0-9]/, "", value); if (value != "") sum += value } END { if (NR > 0) print sum }'
+  )"
+  if [[ -n "$count" ]]; then
+    echo "$count"
+    return
+  fi
+
+  if [[ "$TOOL" == "warp" ]]; then
+    count="$(rg -c 'warp: <ERROR>' "$log_file" || true)"
+  else
+    count="$(rg -ci '(^|[[:space:]])(error|failed|failure)(:|[[:space:]])' "$log_file" || true)"
+  fi
+  echo "${count:-0}"
+}
+
 run_one() {
   local size="$1"
   local log_file="$OUT_DIR/${TOOL}_${size}.log"
@@ -279,29 +300,34 @@ run_one() {
     fi
   fi
 
-  if [[ "$TOOL" == "warp" ]]; then
+  local metrics throughput reqps latency error_count
+  if [[ "$DRY_RUN" != "true" ]]; then
+    metrics="$(collect_metrics "$log_file")"
+    throughput="$(echo "$metrics" | cut -d',' -f1)"
+    reqps="$(echo "$metrics" | cut -d',' -f2)"
+    latency="$(echo "$metrics" | cut -d',' -f3)"
+    error_count="$(collect_error_count "$log_file")"
     # Warp may still exit with code 0 even when it prints runtime failures.
-    # Treat explicit error lines as failed runs to keep summary.csv reliable.
-    if rg -q 'warp: <ERROR>' "$log_file"; then
+    if [[ "$TOOL" == "warp" && "$error_count" != "0" ]]; then
       status="failed"
     fi
+  else
+    throughput="N/A"
+    reqps="N/A"
+    latency="N/A"
+    error_count="N/A"
   fi
 
-  local metrics throughput reqps latency
-  metrics="$(collect_metrics "$log_file")"
-  throughput="$(echo "$metrics" | cut -d',' -f1)"
-  reqps="$(echo "$metrics" | cut -d',' -f2)"
-  latency="$(echo "$metrics" | cut -d',' -f3)"
-
-  echo "$size,$TOOL,$CONCURRENCY,$status,$throughput,$reqps,$latency,$log_file" >> "$SUMMARY_CSV"
+  echo "$size,$TOOL,$CONCURRENCY,$status,$throughput,$reqps,$latency,$error_count,$log_file" >> "$SUMMARY_CSV"
 }
 
 main() {
   parse_args "$@"
   validate_args
   resolve_bucket
-  require_cmd rg
   if [[ "$DRY_RUN" != "true" ]]; then
+    require_cmd rg
+    require_cmd awk
     if [[ "$TOOL" == "warp" ]]; then
       require_cmd "$WARP_BIN"
     else