voonhous commented on PR #18201:
URL: https://github.com/apache/hudi/pull/18201#issuecomment-5633539153

   Where each idea from this PR ended up, after the CI work under #19524:
   
   | this PR | outcome | where |
   |---|---|---|
   | more surefire forks, per-fork port offsets via `-Dsurefire.forkNumber` | 
landed as `forkCount` `0.5C` (two forks on a 4-core runner, one on Azure's 
2-core agents), with the two fixed test ports picked per JVM instead of offset 
| #19892 |
   | shuffle partitions 4 to 2, adaptive execution off | local A/B was inside 
the noise; being measured on CI on its own branch, DNM PR to follow | #19524 |
   | dedicated test-hudi-common job | not taken: one more job and a 3-minute 
build for no wall-clock gain (comment above) | - |
   | shared SparkSession per class for TestCOWDataSource, TestMORDataSource, 
TestHoodieFileSystemViews | measured at zero on its own, context churn is under 
0.1 s per test; kept locally as the prerequisite for concurrency | - |
   | junit-platform.properties, `@Execution(CONCURRENT)` on the six index 
tests, scalatest `parallel=true` | deferred, reasons in the comment above | - |
   
   Related PRs from the same effort: #19890 flattens the job graph (one queue 
hop per run instead of two), #19892 adds the forks, splits 
TestHoodieDeltaStreamer into four classes and reshards the Spark jobs by 
measured time (longest job 77 to 41 min), #19905 restores compiled modules from 
a build cache so a PR only compiles what it touched.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to