cpugputpu created FLINK-16453:
---------------------------------
Summary: A test failure in KafkaTest
Key: FLINK-16453
URL:
https://issues.apache.org/jira/browse/FLINK-16453 Project: Flink
Issue Type: Bug
Components: Connectors / Kafka
Reporter: cpugputpu
The tests in _org.apache.flink.table.descriptors.KafkaTest#testValidation_ may fail. The unexpected behaviour is presented as follows:
testValidation(org.apache.flink.table.descriptors.KafkaTest)
java.lang.AssertionError:
expected:<{connector.property-version=1, connector.startup-mode=specific-offsets,
connector.properties.kafka.stuff=42, connector.properties.zookeeper.stuff=12,
connector.type=kafka, connector.specific-offsets=partition:0,offset:42;partition:1,offset:300,
connector.topic=MyTable, connector.version=0.11}>
but was:<{connector.property-version=1, connector.startup-mode=specific-offsets,
connector.properties.kafka.stuff=42,
connector.properties.zookeeper.stuff=12, connector.type=kafka, connector.specific-offsets=partition:1,
offset:300;partition:0,offset:42, connector.topic=MyTable, connector.version=0.11}>
The root cause of this order issue is regarding two HashMap variables initialized:
this.specificOffsets = new HashMap<>(); (flink-connectors/flink-connector-kafka-base/src/main/java/org/apache/flink/table/descriptors/Kafka.java)
final Map<Integer, Long> offsets = new HashMap<>();(flink-connectors/flink-connector-kafka-base/src/test/java/org/apache/flink/table/descriptors/KafkaTest.java)
And the iteration of HashMap is here:
for (Map.Entry<Integer, Long> specificOffset : specificOffsets.entrySet()) (in function _toConnectorProperties_ __ at _Kafka.java_)
The specification about HashMap says that "this class makes no guarantees as to the order of the map; in particular, it does not guarantee that the order will remain constant over time". The documentation is here for your reference:
https://docs.oracle.com/javase/8/docs/api/java/util/HashMap.html
If I change the two HashMap vairables into LinkedHashMap,the failure will be removed, making the test more stable. If you come up with better fix, please be free to discuss with me.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)