diff --git a/core/database_arango.py b/core/database_arango.py
index cb8efc867..ceaa28718 100644
--- a/core/database_arango.py
+++ b/core/database_arango.py
@@ -195,7 +195,12 @@ def connect(
             try:
                 yeti_db = sys_db.has_database(database)
                 break
-            except requests.exceptions.ConnectionError as e:
+            # python-arango's own host-resolver raises the builtin
+            # ConnectionError when it exhausts its internal retries, rather
+            # than propagating requests' ConnectionError -- catch both, since
+            # which one surfaces depends on where in the request the failure
+            # originates.
+            except (ConnectionError, requests.exceptions.ConnectionError) as e:
                 logging.error("Connection error: {0:s}".format(str(e)))
                 logging.error("Retrying in 5 seconds...")
                 time.sleep(5)
diff --git a/tests/core_tests/database_arango.py b/tests/core_tests/database_arango.py
new file mode 100644
index 000000000..807e5ab84
--- /dev/null
+++ b/tests/core_tests/database_arango.py
@@ -0,0 +1,18 @@
+import unittest
+from unittest import mock
+
+from core import database_arango
+
+
+class ConnectRetryTest(unittest.TestCase):
+    def test_connect_retries_and_exits_cleanly_on_unreachable_host(self) -> None:
+        """connect() must retry (not crash with an unhandled exception) when
+        ArangoDB is unreachable, and exit cleanly once it exhausts its
+        retries. python-arango's host resolver wraps a failed connection
+        attempt in the builtin ConnectionError rather than
+        requests.exceptions.ConnectionError, so the retry loop's except
+        clause needs to catch both."""
+        db = database_arango.ArangoDatabase()
+        with mock.patch("core.database_arango.time.sleep"):
+            with self.assertRaises(SystemExit):
+                db.connect(host="127.0.0.1", port=1, username="root", password="")