[SPARK-45336][TEST] Use Oracle Database 23 Free for integration and testing

[LucaCanali]

What changes were proposed in this pull request?

This proposes to update the Docker image used for integration tests and builds to Oracle Database 23c Free.

Why are the changes needed?

The Docker image used for integration tests and builds currently uses Oracle XE version 21.3.0. Oracle 21 support ends in April 2024. The latest Oracle release is 23c, it is a long-term release supported till 2032. With Oracle 23c, Oracle has changed the name of the free version of its database, from Oracle XE (Express Edition) to Oracle Database Free.

Does this PR introduce any user-facing change?

No

How was this patch tested?

Existing test infrastructure.

Was this patch authored or co-authored using generative AI tooling?

No

[HyukjinKwon]

Seems making sense to me.

[dongjoon-hyun]

+1, LGTM. Thank you, @LucaCanali and @HyukjinKwon .
Merged to master for Apache Spark 4.0.0.

[dongjoon-hyun]

Although this pass the PR builder, this seems to fail in master CI environment.

• https://github.com/apache/spark/actions/runs/6448866897/job/17506605464

[info] OracleIntegrationSuite:
[info] org.apache.spark.sql.jdbc.v2.OracleIntegrationSuite *** ABORTED *** (7 minutes, 55 seconds)
[info]   The code passed to eventually never returned normally. Attempted 410 times over 7.016186794716666 minutes. Last failure message: ORA-12541: Cannot connect. No listener at host 10.1.0.75 port 35471. (CONNECTION_ID=flPajzZKRBul4914CIPekw==). (DockerJDBCIntegrationSuite.scala:166)

[dongjoon-hyun]

After waiting until the current test finishes, let me re-trigger the master branch CI.

[dongjoon-hyun]

It passed at the second try. So, we are good~

[LucaCanali]

Thank you @dongjoon-hyun

[LuciferYang]

Should we backport this patch to branch-3.x? Otherwise, their daily tests will use gvenzl/oracle-free:23.3, but the test code in branch-3.x still uses gvenzl/oracle-xe:21.3.0.

It seems that the OracleIntegrationSuite in the Docker integration tests task of the daily tests for branch-3.x has failed recently.

• branch-3.3: https://github.com/apache/spark/actions/runs/6609791712/job/17950549755
• branch-3.4: https://github.com/apache/spark/actions/runs/6611049884/job/17954225189
• branch-3.5: https://github.com/apache/spark/actions/runs/6612344747/job/17958021656

[LuciferYang]

Or can we use a solution similar to #43306?

For example, for branch-3.x, explicitly add the environment variable ORACLE_DOCKER_IMAGE_NAME like

envs: >-
        {
          "SCALA_PROFILE": "scala2.13",
          "PYTHON_TO_TEST": "",
          "ORACLE_DOCKER_IMAGE_NAME": "gvenzl/oracle-xe:21.3.0"
        }

[dongjoon-hyun]

Backporting would be better for all downstream, @LuciferYang . Feel free to backport.

[LuciferYang]

Or can we use a solution similar to #43306?

For example, for branch-3.x, explicitly add the environment variable ORACLE_DOCKER_IMAGE_NAME like

envs: >-
        {
          "SCALA_PROFILE": "scala2.13",
          "PYTHON_TO_TEST": "",
          "ORACLE_DOCKER_IMAGE_NAME": "gvenzl/oracle-xe:21.3.0"
        }

Hmm... Are we sure that this method can't override this environment variable? If possible, this solution would be relatively simple ... Can you help to confirm this? @HyukjinKwon Thanks

[LuciferYang]

Or can we use a solution similar to #43306?

For example, for branch-3.x, explicitly add the environment variable ORACLE_DOCKER_IMAGE_NAME like

envs: >-
        {
          "SCALA_PROFILE": "scala2.13",
          "PYTHON_TO_TEST": "",
          "ORACLE_DOCKER_IMAGE_NAME": "gvenzl/oracle-xe:21.3.0"
        }

Let me see if I can test this in my local repository.