clickhouse: account for dropped KeeperMap tables

If the table was dropped the table path still exists but the data ZNode
still exists.  This leads to an error in backup

astacus/coordinator/plugins/clickhouse/steps.py

@@ -204,7 +204,12 @@ async def run_step(self, cluster: Cluster, context: StepsContext) -> Sequence[Ke
             for child in children:
                 keeper_map_table_path = os.path.join(self.keeper_map_path_prefix, child)
                 data_path = os.path.join(keeper_map_table_path, "data")
-                data = await connection.get_children_with_data(data_path)
+                try:
+                    data = await connection.get_children_with_data(data_path)
+                except NoNodeError:
+                    logger.warning("ZNode %s is missing, table was dropped.  Skipping", data_path)
+                    continue
+
                 tables.append(
                     KeeperMapTable(
                         name=child,

tests/integration/coordinator/plugins/clickhouse/test_plugin.py

@@ -316,6 +316,12 @@ async def setup_cluster_content(clients: Sequence[HttpClickHouseClient], use_nam
     await clients[0].execute(b"INSERT INTO default.memory VALUES (123, 'foo')")
     await clients[0].execute(b"CREATE FUNCTION `linear_equation_\x80` AS (x, k, b) -> k*x + b")
     if await is_engine_available(clients[0], TableEngine.KeeperMap):
+        # add and drop a table with KeeperMap engine.  This leaves some metadata in ZooKeeper that we want to ignore.
+        await clients[0].execute(
+            b"CREATE TABLE default.keeper_map_dropped (key UInt32, value String) ENGINE = KeeperMap('keeper_map_dropped') PRIMARY KEY key"
+        )
+        await clients[0].execute(b"DROP TABLE default.keeper_map_dropped")
+        # add a table to backup
         await clients[0].execute(
             b"CREATE TABLE default.keeper_map (key UInt32, value String) ENGINE = KeeperMap('keeper_map') PRIMARY KEY key"
         )