[ 
https://issues.apache.org/jira/browse/FLINK-37937?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Etienne Chauchot reassigned FLINK-37937:
----------------------------------------

    Assignee: Etienne Chauchot

> Test Cassandra source in real cluster conditions to better test the split.
> --------------------------------------------------------------------------
>
>                 Key: FLINK-37937
>                 URL: https://issues.apache.org/jira/browse/FLINK-37937
>             Project: Flink
>          Issue Type: Technical Debt
>          Components: Connectors / Cassandra
>            Reporter: Etienne Chauchot
>            Assignee: Etienne Chauchot
>            Priority: Major
>
> A bug in split calculation (ring fraction calculation) was uncovered by [this 
> PR|https://github.com/apache/flink-connector-cassandra/pull/24].  The 
> existing split tests are run on an embedded Cassandra cluster with only one 
> node. This leads to having ringFraction always equal to 1 (the single node 
> hosts 100% of the data) during the tests. This masks the bug. If the 
> available resources allow it, it would be interesting to test splits on an 
> embedded cluster of 2 nodes with a testContainers code similar to below:
> {code:java}
> Network network = Network.newNetwork();
> CassandraContainer<?> cassandra1 = new CassandraContainer<>("cassandra:4.0")
>     .withNetwork(network)
>     .withNetworkAliases("cassandra1");
> CassandraContainer<?> cassandra2 = new CassandraContainer<>("cassandra:4.0")
>     .withNetwork(network)
>     .withEnv("CASSANDRA_SEEDS", "cassandra1");
> cassandra1.start();
> cassandra2.start();
> {code}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to