diff --git a/test/cluster/test_describe.py b/test/cluster/test_describe.py
index 46a693efe7..ed56869b1a 100644
--- a/test/cluster/test_describe.py
+++ b/test/cluster/test_describe.py
@@ -8,6 +8,12 @@ import asyncio
 import pytest
 from test.cluster.util import new_test_keyspace, new_test_table
 from test.pylib.manager_client import ManagerClient
+from test.pylib.util import wait_for
+from cassandra.connection import UnixSocketEndPoint
+from cassandra.policies import WhiteListRoundRobinPolicy
+from test.cluster.conftest import cluster_con
+from time import time
+import os
 
 # The following test verifies that Scylla avoids making an oversized allocation
 # when generating a large create statement when performing a DESCRIBE statement.
@@ -46,3 +52,41 @@ async def test_large_create_statement(manager: ManagerClient):
 
             matches = await log.grep("oversized allocation", from_mark=marker)
             assert len(matches) == 0
+
+@pytest.mark.parametrize("mode", ["normal", "maintenance"])
+@pytest.mark.asyncio
+async def test_describe_cluster_sanity(manager: ManagerClient, mode: str):
+    """
+    Parametrized test that DESCRIBE CLUSTER returns correct cluster information
+    in both normal and maintenance modes.
+
+    This test verifies that cluster metadata from gossiper is properly initialized
+    and the cluster name is consistent with system.local in both:
+    - normal mode: standard cluster operation
+    - maintenance mode: node isolated from the cluster
+    """
+
+    if mode == "normal":
+        await manager.server_add()
+        cql = manager.get_cql()
+    else:  # maintenance mode
+        srv = await manager.server_add(config={"maintenance_mode": True}, connect_driver=False)
+        maintenance_socket_path = await manager.server_get_maintenance_socket_path(srv.server_id)
+        async def socket_exists():
+            return True if os.path.exists(maintenance_socket_path) else None
+        await wait_for(socket_exists, time() + 30)
+        socket_endpoint = UnixSocketEndPoint(maintenance_socket_path)
+        cluster = cluster_con([socket_endpoint], load_balancing_policy=WhiteListRoundRobinPolicy([socket_endpoint]))
+        cql = cluster.connect()
+
+    try:
+        system_local_results = await cql.run_async("SELECT cluster_name FROM system.local")
+        assert system_local_results[0].cluster_name != ""  # sanity check
+
+        describe_results = await cql.run_async("DESCRIBE CLUSTER")
+        assert describe_results[0].partitioner == 'org.apache.cassandra.dht.Murmur3Partitioner'
+        assert describe_results[0].snitch == 'org.apache.cassandra.locator.SimpleSnitch'
+        assert describe_results[0].cluster == system_local_results[0].cluster_name
+    finally:
+        if mode == "maintenance":
+            cluster.shutdown()