(zeppelin) branch master updated: [ZEPPELIN-6454] Stabilize and re-enable flaky PyFlink streaming SQL tests

[yo...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

voidmatcha pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/zeppelin.git


The following commit(s) were added to refs/heads/master by this push:
     new a865c949ff [ZEPPELIN-6454] Stabilize and re-enable flaky PyFlink streaming SQL tests
a865c949ff is described below

commit a865c949ff0841dc09bae28026a550af05699638
Author: Lee SuJung <15...@users.noreply.github.com>
AuthorDate: Thu Oct 15 20:50:17 2026 +0900

    [ZEPPELIN-6454] Stabilize and re-enable flaky PyFlink streaming SQL tests
    
    ### What is this PR for?
    
    Re-enables `IPyFlinkInterpreterTest.testCancelStreamSql`. The two other disabled
    cases are a different situation and are left alone.
    
    Why this one can be re-enabled:
    
    - `PyFlinkInterpreterTest.testCancelStreamSql` calls the same static
      `IPyFlinkInterpreterTest.testCancelStreamSql` helper and is not disabled, so it
      runs on every build. The cancel logic is already covered in CI; only the
      IPython variant was left off.
    - The disabling commit `62a892792` (2021-03-03) also raised the Flink log levels
      from WARN to INFO. No issue or PR is linked to it.
    - Eight days later `ZEPPELIN-5284` landed: on Flink 1.12 the interpreter updated
      a copied `Configuration`, so savepoints and checkpoints did not work. Same area
      as the hotfix, and fixed since.
    
    ### What type of PR is it?
    Improvement
    
    ### Todos
    * [x] Remove `<at>Disabled` from the IPython variant
    * [x] Confirm it passes locally
    * [x] Confirm on CI, where the runner is slower than a dev machine
    
    ### What is the Jira issue?
    * [ZEPPELIN-6454](https://issues.apache.org/jira/browse/ZEPPELIN-6454)
    
    ### How should this be tested?
    
    ```
    ./mvnw clean install -pl flink/flink-scala-2.12 --am -Pflink-1.20 -DskipTests
    ./mvnw test -pl flink/flink-scala-2.12 -Pflink-1.20 -Dtest=IPyFlinkInterpreterTest#testCancelStreamSql
    ```
    
    Two prerequisites:
    
    - `clean` is needed on the install step. Without it the Scala sources are skipped
      and the test fails to compile with `FlinkScalaInterpreter cannot be resolved to
      a type`.
    - A PyFlink environment has to be on `PATH`, since the test hardcodes
      `zeppelin.pyflink.python=python` (I used a venv with `apache-flink==1.20.0`).
    
    Ran four times, all passing:
    
    ```
    Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 104.7 s
    Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 102.9 s
    Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 104.0 s
    Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 102.8 s
    ```
    
    CI job: `flink-test-and-flink-integration-test` in `core.yml` (java 17 /
    flink-1.20).
    
    Two limits:
    
    - The flakiness did not reproduce. Besides the four runs above, 12 earlier runs
      also passed (8 unloaded, 1 under heavy load, 3 under moderate load). Load moved
      the duration from 108s to 114s, so the helper's 20-second wait was never close
      to insufficient here. A slower CI runner is the case I cannot reproduce
      locally.
    - Historical CI logs are gone. The oldest retained workflow run in this repo is
      from 2025-07-28, years after the test was disabled.
    
    If it does fail on CI, that failure is the first concrete evidence about this
    test, and it can go back to being disabled with that evidence attached.
    
    ### Why the savepoint tests stay disabled
    
    - `testResumeStreamSqlFromSavePoint` (both variants) was committed in
      `3cf32a093` (2020-04-02) with `// <at>Test` already commented out, so it has never
      run in CI.
    - That is not stabilizing a flaky test but making a never-verified test work, and
      it needs its own investigation.
    
    ### Questions:
    * Does the license files need to update? No
    * Is there breaking changes for older versions? No
    * Does this needs documentation? No
    
    
    Closes #5571 from xhaktm00/ZEPPELIN-6454.
    
    Signed-off-by: YONGJAE LEE <de...@gmail.com>
---
 .../src/test/java/org/apache/zeppelin/flink/IPyFlinkInterpreterTest.java | 1 -
 1 file changed, 1 deletion(-)

diff --git a/flink/flink-scala-2.12/src/test/java/org/apache/zeppelin/flink/IPyFlinkInterpreterTest.java b/flink/flink-scala-2.12/src/test/java/org/apache/zeppelin/flink/IPyFlinkInterpreterTest.java
index 3fcc876736..ecb42a204b 100644
--- a/flink/flink-scala-2.12/src/test/java/org/apache/zeppelin/flink/IPyFlinkInterpreterTest.java
+++ b/flink/flink-scala-2.12/src/test/java/org/apache/zeppelin/flink/IPyFlinkInterpreterTest.java
@@ -150,7 +150,6 @@ public class IPyFlinkInterpreterTest extends IPythonInterpreterTest {
   }
 
   @Test
-  @Disabled("(zjffdu) flaky test")
   void testCancelStreamSql()
       throws InterpreterException, IOException, TimeoutException, InterruptedException {
     testCancelStreamSql(interpreter, flinkScalaInterpreter);