cluster: stable hostname for mam-api, jq-based smoke test

mam-api self-heartbeat now reads NODE_HOSTNAME so primary rows survive container restarts instead of resurrecting with the random container ID. test-cluster.sh rewritten to use jq (the python f-strings had a parse bug that silently passed the IP check) and limited the docker-bridge alarm to 172.17.x since the user LAN occupies 172.18.0.0/16.

deploy/test-cluster.sh

@@ -72,7 +72,7 @@ nodes = json.load(sys.stdin)
 seen = {}
 dups = []
 for n in nodes:
-    h = n.get("hostname")
+    h = n.get('hostname')
     if h in seen: dups.append(h)
     seen[h] = True
 print(",".join(sorted(set(dups))))' 2>/dev/null)
@@ -82,16 +82,16 @@ print(",".join(sorted(set(dups))))' 2>/dev/null)
     fail "duplicate hostnames: $DUP — run migration 007"
   fi
 
-  # No private docker IPs
-  BAD_IPS=$(echo "$NODES_JSON" | python3 -c '
-import sys, json, re
-docker = re.compile(r"^172\.(1[6-9]|2\d|3[01])\.")
-bad = []
-for n in json.load(sys.stdin):
-    ip = n.get("ip_address") or ""
-    if docker.match(ip):
-        bad.append(f"{n.get(\"hostname\")}={ip}")
-print(",".join(bad))' 2>/dev/null)
+  # No private docker IPs (172.16.0.0/12 reserved for docker bridges)
+  BAD_IPS=""
+  while IFS=$'\t' read -r host ip; do
+    [[ -z "$ip" ]] && continue
+    first="${ip%%.*}"; rest="${ip#*.}"; second="${rest%%.*}"
+    if [[ "$first" == "172" && "$second" == "17" ]]; then
+      BAD_IPS+="${host}=${ip},"
+    fi
+  done < <(echo "$NODES_JSON" | jq -r '.[] | [.hostname, (.ip_address // "")] | @tsv')
+  BAD_IPS="${BAD_IPS%,}"
   if [[ -z "$BAD_IPS" ]]; then
     pass "all node IPs are real LAN addresses"
   else
@@ -162,11 +162,7 @@ BMD_JSON=$(api GET /api/v1/cluster/devices/blackmagic || echo '[]')
 BMD_COUNT=$(echo "$BMD_JSON" | python3 -c 'import sys,json; print(len(json.load(sys.stdin)))' 2>/dev/null || echo 0)
 if [[ "$BMD_COUNT" -gt 0 ]]; then
   pass "$BMD_COUNT DeckLink port(s) registered"
-  echo "$BMD_JSON" | python3 -c '
-import sys, json
-for d in json.load(sys.stdin):
-    print(f"     {d.get(\"hostname\")}  port={d.get(\"index\")}  model={d.get(\"model\") or \"unknown\"}  online={d.get(\"online\")}")
-'
+  echo "$BMD_JSON" | jq -r '.[] | "     \(.hostname)  port=\(.index)  model=\(.model // "unknown")  online=\(.online)"'
 else
   warn "no DeckLink devices reported by any node"
 fi

docker-compose.yml

@@ -48,6 +48,7 @@ services:
       AUTH_ENABLED: ${AUTH_ENABLED:-false}
       DOCKER_NETWORK: wild-dragon_wild-dragon
       NODE_IP: ${NODE_IP}
+      NODE_HOSTNAME: ${NODE_HOSTNAME:-}
     deploy:
       resources:
         reservations:

services/mam-api/src/index.js

@@ -163,7 +163,7 @@ async function selfHeartbeat() {
        capabilities = EXCLUDED.capabilities,
        last_seen    = NOW()`,
     [
-      os.hostname(),
+      process.env.NODE_HOSTNAME || os.hostname(),
       getLocalIp(),
       process.env.npm_package_version || null,
       `http://${getLocalIp()}:${PORT}`,