monitoring: harden grafana user dedupe

services/monitoring/grafana-user-dedupe-job.yaml

@@ -2,7 +2,7 @@
 apiVersion: batch/v1
 kind: Job
 metadata:
-  name: grafana-user-dedupe-api-v5
+  name: grafana-user-dedupe-api-v6
   namespace: monitoring
 spec:
   backoffLimit: 1
@@ -60,35 +60,66 @@ spec:
               import json
               import os
               import urllib.parse
+              import urllib.error
               import urllib.request
 
               grafana_url = os.environ["GRAFANA_URL"].rstrip("/")
               user = os.environ["GRAFANA_USER"]
               password = os.environ["GRAFANA_PASSWORD"]
-              emails = [e.strip() for e in os.environ["GRAFANA_DEDUPE_EMAILS"].split(",") if e.strip()]
+              lookups = [e.strip() for e in os.environ["GRAFANA_DEDUPE_EMAILS"].split(",") if e.strip()]
 
               token = base64.b64encode(f"{user}:{password}".encode("utf-8")).decode("utf-8")
               headers = {"Authorization": f"Basic {token}"}
 
               def request(method: str, url: str):
                   req = urllib.request.Request(url, headers=headers, method=method)
-                  with urllib.request.urlopen(req, timeout=10) as resp:
-                      return resp.read()
-
-              for email in emails:
-                  lookup_url = f"{grafana_url}/api/users/lookup?loginOrEmail={urllib.parse.quote(email)}"
                   try:
-                      payload = json.loads(request("GET", lookup_url))
+                      with urllib.request.urlopen(req, timeout=10) as resp:
-                  except Exception:
+                          return resp.status, resp.read()
-                      print(f"no grafana user found for {email}")
+                  except urllib.error.HTTPError as err:
+                      body = err.read()
+                      return err.code, body
+
+              for _ in range(60):
+                  status, _ = request("GET", f"{grafana_url}/api/health")
+                  if status == 200:
+                      break
+              else:
+                  raise SystemExit("Grafana API did not become ready in time")
+
+              for lookup in lookups:
+                  search_url = f"{grafana_url}/api/users/search?query={urllib.parse.quote(lookup)}"
+                  status, body = request("GET", search_url)
+                  if status != 200:
+                      print(f"search failed for {lookup}: status={status} body={body.decode('utf-8', errors='ignore')}")
                       continue
-                  user_id = payload.get("id")
+                  payload = json.loads(body)
+                  users = payload.get("users", [])
+                  matches = [
+                      user
+                      for user in users
+                      if user.get("email", "").lower() == lookup.lower()
+                      or user.get("login", "").lower() == lookup.lower()
+                  ]
+                  if not matches:
+                      print(f"no grafana user found for {lookup}")
+                      continue
+                  for user in matches:
+                      user_id = user.get("id")
                       if not user_id:
-                      print(f"no grafana user found for {email}")
                           continue
-                  print(f"deleting grafana user {user_id} ({email})")
+                      print(f"deleting grafana user {user_id} ({user.get('email')})")
                       delete_url = f"{grafana_url}/api/admin/users/{user_id}"
-                  request("DELETE", delete_url)
+                      del_status, del_body = request("DELETE", delete_url)
+                      if del_status not in (200, 202, 204):
+                          print(
+                              "delete failed for",
+                              user_id,
+                              "status",
+                              del_status,
+                              "body",
+                              del_body.decode("utf-8", errors="ignore"),
+                          )
               PY
               echo "done"
           env: