[CASSANDRA-19251] [Analytics] Speed up integration tests

• Type: Improvement
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: NA
• Component/s: Analytics Library
• Change Category: Quality Assurance
• Complexity: Challenging
• Platform: All

Source Control Link

https://github.com/apache/cassandra-analytics/commit/550bdfa1c6082537e2cfb93449128a61dbe3a1fb

Test and Documentation Plan

Reworked integration tests

• Assignee: Francisco Guerrero
• Reporter: Francisco Guerrero

Authors

Francisco Guerrero

Reviewers

Yifan Cai

• Created: 05/Jan/24 17:36
• Updated: 10/Jan/24 13:46
• Resolved: 10/Jan/24 13:45

Description

We can speed up the integration tests by reusing the in-jvm dtest cluster in a test class when possible. The proposal has the following pros and cons.

Pros:

1. Ability to iterate faster by running tests faster
2. Reduce the resources needed to run tests
3. Reduce flakiness due to resource allocation issues
4. Remove necessity to spin up a new process for bulk writes
5. Adds the ability to use parameterized tests (for example different CLs for the same test)

Cons:

1. Potentially having to add new test classes
2. No longer be able to use the CassandraIntegrationTest annotation
3. Can only test a single Cassandra version at a time

In terms of resources, the proposed approach brings huge improvements by reducing the current runtime of about 1 hour down to 20 minutes. Additionally, we reduce the resource requirements for the integration tests, and add the ability to iterate faster. Reduces flakiness which helps surface real issues in the tests.

Issue Links

• links to: GitHub Pull Request #29