Add cdp-latency-check.sh and box cdp-latency action

Proper script replacing the inline-SSH latency monitor. Probes each relay /json/version via pinned ports from netvm-names.sh, outputs name:latency_ms:code per node. box-ctl.py cdp-latency runs it and returns JSON.

bin/box-ctl.py

@@ -697,6 +697,35 @@ def act_relay_health():
     out(all_healthy, relays=results, healthy=all_healthy)
 
 
+
+def act_cdp_latency():
+    """Run cdp-latency-check.sh and return per-node latency JSON."""
+    audit("cdp-latency")
+    script = BIN / "cdp-latency-check.sh"
+    try:
+        r = subprocess.run([str(script)], capture_output=True, text=True, timeout=120)
+    except subprocess.TimeoutExpired:
+        fail("LATENCY_TIMEOUT", "cdp-latency-check.sh timed out")
+    nodes = []
+    for line in r.stdout.strip().split("\n"):
+        parts = line.strip().split(":")
+        if len(parts) != 3:
+            continue
+        name, lat, code = parts
+        entry = {"node": name.strip(), "http_code": code.strip()}
+        if lat.strip() == "FAIL":
+            entry["ok"] = False
+            entry["latency_ms"] = None
+        else:
+            try:
+                entry["latency_ms"] = int(lat.strip())
+            except ValueError:
+                entry["latency_ms"] = None
+            entry["ok"] = code.strip() == "200"
+        nodes.append(entry)
+    all_ok = bool(nodes) and all(n["ok"] for n in nodes)
+    out(all_ok, nodes=nodes)
+
 def act_chrome_errors():
     """Run chrome-error-scan.sh and return per-profile error counts as JSON."""
     audit("chrome-errors")
@@ -1101,6 +1130,7 @@ fleet:
   chrome-errors         per-profile chrome FATAL/crash counts (new since watermark)
   watchdog-alerts
   relay-health
+  cdp-latency
 
 timer actions:
   timer-list
@@ -1333,6 +1363,8 @@ def main(argv):
         act_watchdog_alerts()
     elif action == "relay-health":
         act_relay_health()
+    elif action == "cdp-latency":
+        act_cdp_latency()
     elif action == "chrome-errors":
         if rest:
             fail("BAD_ARGS", "usage: chrome-errors")

bin/cdp-latency-check.sh

@@ -0,0 +1,33 @@
+#!/bin/bash
+# cdp-latency-check.sh — measure CDP relay latency per node.
+#
+# Probes each node's host-side relay (PEER_IP:CDP_PORT from netvm-names.sh,
+# which pins the registry ports) at /json/version and prints one line per
+# node:
+#   name:latency_ms:code
+# On failure (timeout / connection refused / non-200-ish transport error):
+#   name:FAIL:000
+#
+# Self-contained: only needs netvm-names.sh in the same directory and curl.
+set -u
+
+BIN_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+# shellcheck disable=SC1091
+source "$BIN_DIR/netvm-names.sh"
+
+TIMEOUT_S=10
+
+for node in muse pip 646 opm; do
+  netvm_names "$node"
+  url="http://${PEER_IP}:${CDP_PORT}/json/version"
+  probe=$(curl -s -m "$TIMEOUT_S" -o /dev/null -w "%{time_total} %{http_code}" "$url" 2>/dev/null)
+  rc=$?
+  secs=$(printf '%s' "$probe" | awk '{print $1}')
+  code=$(printf '%s' "$probe" | awk '{print $2}')
+  if [ "$rc" -ne 0 ] || [ -z "$code" ] || [ "$code" = "000" ]; then
+    echo "${node}:FAIL:000"
+    continue
+  fi
+  ms=$(awk -v s="$secs" 'BEGIN { printf "%d", (s + 0) * 1000 }')
+  echo "${node}:${ms}:${code}"
+done