KAFKA-15378: fix streams upgrade system test (#14539)

Fixing bad test setup. We tried to fix an upgrade bug for FK-joins in 3.1 release, but it later turned out that the PR was not sufficient to fix it. We finally fixed in 3.4 release.

This PR updates the system test matrix to only test working versions with FK-joins, limited to available test versions.

Reviewers: Guozhang Wang <wangguoz@gmail.com>, Hao Li <hli@confluent.io>, Mickael Maison <mickael.maison@gmail.com>

tests/kafkatest/tests/streams/streams_broker_down_resilience_test.py

@@ -100,7 +100,7 @@ def test_streams_runs_with_broker_down_initially(self, metadata_quorum):
         processor_3 = StreamsBrokerDownResilienceService(self.test_context, self.kafka, configs)
         processor_3.start()
 
-        broker_unavailable_message = "Broker may not be available"
+        broker_unavailable_message = "Node may not be available"
 
         # verify streams instances unable to connect to broker, kept trying
         self.wait_for_verification(processor, broker_unavailable_message, processor.LOG_FILE, 10)

tests/kafkatest/tests/streams/streams_upgrade_test.py

@@ -40,11 +40,13 @@
 metadata_1_versions = [str(LATEST_0_10_0)]
 metadata_2_versions = [str(LATEST_0_10_1), str(LATEST_0_10_2), str(LATEST_0_11_0), str(LATEST_1_0), str(LATEST_1_1),
                        str(LATEST_2_4), str(LATEST_2_5), str(LATEST_2_6), str(LATEST_2_7), str(LATEST_2_8),
-                       str(LATEST_3_0)]
-# upgrading from version (2.4...3.0) is broken and only fixed later in 3.1
-# we cannot test two bounce rolling upgrade because we know it's broken
-# instead we add version 2.4...3.0 to the `metadata_2_versions` upgrade list
-fk_join_versions = [str(LATEST_3_1), str(LATEST_3_2), str(LATEST_3_3)]
+                       str(LATEST_3_0), str(LATEST_3_1), str(LATEST_3_2), str(LATEST_3_3)]
+# upgrading from version (2.4...3.3) is broken and only fixed later in 3.3.3 (unreleased) and 3.4.0
+# -> https://issues.apache.org/jira/browse/KAFKA-14646
+# thus, we cannot test two bounce rolling upgrade because we know it's broken
+# instead we add version 2.4...3.3 to the `metadata_2_versions` upgrade list
+#fk_join_versions = [str(LATEST_3_4)]
+
 
 """
 After each release one should first check that the released version has been uploaded to 
@@ -202,7 +204,7 @@ def test_upgrade_downgrade_brokers(self, from_version, to_version):
     @cluster(num_nodes=6)
     @matrix(from_version=metadata_1_versions, to_version=[str(DEV_VERSION)])
     @matrix(from_version=metadata_2_versions, to_version=[str(DEV_VERSION)])
-    @matrix(from_version=fk_join_versions, to_version=[str(DEV_VERSION)])
+    #@matrix(from_version=fk_join_versions, to_version=[str(DEV_VERSION)])
     def test_rolling_upgrade_with_2_bounces(self, from_version, to_version):
         """
         This test verifies that the cluster successfully upgrades despite changes in the metadata and FK

tests/kafkatest/version.py

@@ -107,8 +107,10 @@ def supports_topic_ids_when_using_zk(self):
         return self >= V_2_8_0
 
     def supports_fk_joins(self):
-        # while we support FK joins since 2.4, rolling upgrade is broken in older versions and only fixed in 3.1
-        return hasattr(self, "version") and self >= V_3_1_2
+        # while we support FK joins since 2.4, rolling upgrade is broken in older versions
+        # it's only fixed in 3.3.3 (unreleased) and 3.4.0
+        # -> https://issues.apache.org/jira/browse/KAFKA-14646
+        return hasattr(self, "version") and self >= V_3_4_0
 
 def get_version(node=None):
     """Return the version attached to the given node.