Merge pull request 'fix: map SSH-discovered nodes by VMID when hostnames are generic' (#105) from stage into master

Reviewed-on: #105

nixos/kubeadm/scripts/discover-inventory-from-ssh.py

@@ -6,7 +6,7 @@ import json
 import os
 import subprocess
 import sys
-from typing import Dict, Set
+from typing import Dict, Set, Tuple
 
 
 def derive_prefix(payload: dict) -> str:
@@ -25,7 +25,7 @@ def derive_prefix(payload: dict) -> str:
     return "10.27.27"
 
 
-def ssh_hostname(ip: str, users: list[str], key_path: str, timeout_sec: int) -> tuple[str, str] | None:
+def ssh_probe(ip: str, users: list[str], key_path: str, timeout_sec: int) -> Tuple[str, str, str] | None:
     cmd_tail = [
         "-o",
         "BatchMode=yes",
@@ -39,13 +39,23 @@ def ssh_hostname(ip: str, users: list[str], key_path: str, timeout_sec: int) ->
         key_path,
     ]
     for user in users:
-        cmd = ["ssh", *cmd_tail, f"{user}@{ip}", "hostnamectl --static 2>/dev/null || hostname"]
+        cmd = [
+            "ssh",
+            *cmd_tail,
+            f"{user}@{ip}",
+            "hn=$(hostnamectl --static 2>/dev/null || hostname); serial=$(cat /sys/class/dmi/id/product_serial 2>/dev/null || true); printf '%s|%s\n' \"$hn\" \"$serial\"",
+        ]
         try:
             out = subprocess.check_output(cmd, stderr=subprocess.DEVNULL, text=True, timeout=timeout_sec + 2).strip()
         except Exception:
             continue
         if out:
-            return out.splitlines()[0].strip(), ip
+            line = out.splitlines()[0].strip()
+            if "|" in line:
+                host, serial = line.split("|", 1)
+            else:
+                host, serial = line, ""
+            return host.strip(), ip, serial.strip()
     return None
 
 
@@ -89,21 +99,30 @@ def main() -> int:
 
     scan_ips = [str(ipaddress.IPv4Address(f"{prefix}.{i}")) for i in range(start, end + 1)]
     found: Dict[str, str] = {}
+    vmid_to_name: Dict[str, str] = {}
+    for name, vmid in payload.get("control_plane_vm_ids", {}).get("value", {}).items():
+        vmid_to_name[str(vmid)] = name
+    for name, vmid in payload.get("worker_vm_ids", {}).get("value", {}).items():
+        vmid_to_name[str(vmid)] = name
 
     seen_hostnames: Dict[str, str] = {}
 
     def run_pass(pass_timeout: int, pass_workers: int) -> None:
         with concurrent.futures.ThreadPoolExecutor(max_workers=pass_workers) as pool:
-            futures = [pool.submit(ssh_hostname, ip, users, key_path, pass_timeout) for ip in scan_ips]
+            futures = [pool.submit(ssh_probe, ip, users, key_path, pass_timeout) for ip in scan_ips]
             for fut in concurrent.futures.as_completed(futures):
                 result = fut.result()
                 if not result:
                     continue
-                host, ip = result
+                host, ip, serial = result
                 if host not in seen_hostnames:
                     seen_hostnames[host] = ip
                 if host in target_names and host not in found:
                     found[host] = ip
+                elif serial in vmid_to_name:
+                    inferred = vmid_to_name[serial]
+                    if inferred not in found:
+                        found[inferred] = ip
                 if all(name in found for name in target_names):
                     return