ducky: change the balanced partition criteria

The assessment whether partitions are balanced were done by comparing
the number of partitions in each node with the average target ±20%. This
way went broken with the introduction of partition balancing domains.
The criteria is changed by this commit to make sure that the number of
partitions across noded is levelled within a scope of each domain
separately. Levelled means that min and max # of replicas differ by 1
at most.
Re redpanda-data#7418

tests/rptest/tests/scaling_up_test.py

@@ -27,41 +27,47 @@ class ScalingUpTest(EndToEndTest):
     rebalance_timeout = 120
     group_topic_partitions = 16
 
-    def _replicas_per_node(self):
+    # returns [{common domain topics replicas},{consumer_offsets replicas}]
+    def _replicas_per_domain_per_node(self):
         kafkacat = KafkaCat(self.redpanda)
-        node_replicas = {}
+        node_replicas = [{}, {}]
         md = kafkacat.metadata()
         self.redpanda.logger.debug(f"metadata: {md}")
         for topic in md['topics']:
+            # this index doesn't agree to the numbering in cluster/types.h
+            # but that does not really matter until there will be more domains
+            i = 1 if topic['topic'] == '__consumer_offsets' else 0
             for p in topic['partitions']:
                 for r in p['replicas']:
                     id = r['id']
-                    if id not in node_replicas:
-                        node_replicas[id] = 0
-                    node_replicas[id] += 1
+                    if id not in node_replicas[i]:
+                        node_replicas[i][id] = 0
+                    node_replicas[i][id] += 1
 
         return node_replicas
 
     def wait_for_partitions_rebalanced(self, total_replicas, timeout_sec):
-
-        expected_per_node = total_replicas / len(self.redpanda.started_nodes())
-        expected_range = [0.8 * expected_per_node, 1.2 * expected_per_node]
-
         def partitions_rebalanced():
-            per_node = self._replicas_per_node()
+            per_domain = self._replicas_per_domain_per_node()
             self.redpanda.logger.info(
-                f"replicas per node: {per_node}, expected range: [{expected_range[0]},{expected_range[1]}]"
-            )
-            if len(per_node) < len(self.redpanda.started_nodes()):
-                return False
+                f"replicas per node: common domain: {per_domain[0]}, "
+                f"__consumer_offsets: {per_domain[1]} ")
 
-            replicas = sum(per_node.values())
-            if replicas != total_replicas:
+            # total # of nodes
+            if len(set().union(*(n.keys() for n in per_domain))) < len(
+                    self.redpanda.started_nodes()):
                 return False
 
-            if not all(expected_range[0] <= p[1] <= expected_range[1]
-                       for p in per_node.items()):
+            # total # of replicas
+            if sum(p for n in per_domain
+                   for p in n.values()) != total_replicas:
                 return False
+
+            # replicas balanced condition separately for each domain
+            for per_node in per_domain:
+                if max(per_node.values()) - min(per_node.values()) > 1:
+                    return False
+
             admin = Admin(self.redpanda)
 
             # make sure that all reconfigurations are finished
@@ -184,7 +190,7 @@ def test_on_demand_rebalancing(self, partition_count):
 
         # verify that all new nodes are empty
 
-        per_node = self._replicas_per_node()
+        per_node = self._replicas_per_domain_per_node()
 
         assert len(per_node) == 3