WIP test

tests/rptest/tests/end_to_end.py

@@ -79,6 +79,7 @@ def __init__(self,
 
     def start_redpanda(self,
                        num_nodes=1,
+                       num_started_nodes=None,
                        extra_rp_conf=None,
                        si_settings=None,
                        environment=None,
@@ -122,7 +123,11 @@ def start_redpanda(self,
             self.redpanda._installer.install(self.redpanda.nodes,
                                              version_to_install)
 
-        self.redpanda.start(auto_assign_node_id=new_bootstrap,
+        started_nodes = None
+        if num_started_nodes is not None:
+            started_nodes = self.redpanda.nodes[:num_started_nodes]
+        self.redpanda.start(nodes=started_nodes,
+                            auto_assign_node_id=new_bootstrap,
                             omit_seeds_on_idx_one=not new_bootstrap)
         if version_to_install and install_opts.num_to_upgrade > 0:
             # Perform the upgrade rather than starting each node on the

tests/rptest/tests/partition_balancer_test.py

@@ -1142,3 +1142,62 @@ def test_transfer_controller_leadership(self):
                                          target_id=transfer_to_idx)
 
             self.wait_until_ready()
+
+    @skip_debug_mode
+    @cluster(num_nodes=7, log_allow_list=CHAOS_LOG_ALLOW_LIST)
+    def test_rebalance_finish(self):
+        self.start_redpanda(num_nodes=5,
+                            num_started_nodes=3,
+                            new_bootstrap=True,
+                            extra_rp_conf={
+                                "partition_autobalancing_mode": "node_add",
+                            })
+        self.topic = TopicSpec(partition_count=50)
+        self.client().create_topic(self.topic)
+
+        self.start_producer(1)
+        self.start_consumer(1)
+        self.await_startup()
+
+        seed_nodes = self.redpanda.nodes[:3]
+        self.redpanda.restart_nodes(
+            seed_nodes,
+            auto_assign_node_id=True,
+            omit_seeds_on_idx_one=False,
+            override_cfg_params={"recovery_mode_enabled": True})
+
+        joiner_nodes = self.redpanda.nodes[3:]
+        for node in joiner_nodes:
+            self.redpanda.start_node(node,
+                                     auto_assign_node_id=True,
+                                     omit_seeds_on_idx_one=False)
+        self.redpanda.wait_for_membership(first_start=False)
+
+        admin = Admin(self.redpanda)
+
+        admin.transfer_leadership_to(namespace='redpanda',
+                                     topic='controller',
+                                     partition=0,
+                                     target_id=self.redpanda.node_id(
+                                         joiner_nodes[0]))
+
+        self.wait_until_status(lambda s: s["status"] == "stalled")
+
+        partition_counts = [
+            len(admin.get_partitions(node=n)) for n in joiner_nodes
+        ]
+        self.logger.info(f"FFF1 partition counts: {partition_counts}")
+
+        # restart in normal mode
+        self.redpanda.restart_nodes(seed_nodes, auto_assign_node_id=True)
+        self.redpanda.wait_for_membership(first_start=False)
+
+        self.wait_until_ready()
+
+        partition_counts = [
+            len(admin.get_partitions(node=n)) for n in self.redpanda.nodes
+        ]
+        self.logger.info(f"FFF2 partition counts: {partition_counts}")
+
+        self.run_validation(min_records=100,
+                            consumer_timeout_sec=CONSUMER_TIMEOUT)