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

HyukjinKwon pushed a commit to branch branch-4.1
in repository https://gitbox.apache.org/repos/asf/spark.git


The following commit(s) were added to refs/heads/branch-4.1 by this push:
     new 95c9796438b1 [SPARK-58026][INFRA] Bound SparkR Windows test step to 
avoid silent 2h job timeout
95c9796438b1 is described below

commit 95c9796438b1bcf9cfc8dd7ee21737faef3d710b
Author: Hyukjin Kwon <[email protected]>
AuthorDate: Wed Jul 8 13:17:47 2026 +0900

    [SPARK-58026][INFRA] Bound SparkR Windows test step to avoid silent 2h job 
timeout
    
    ### What changes were proposed in this pull request?
    
    Split out from the flaky-test deflake work; verified in fork CI (see below).
    
    The scheduled `build_sparkr_window` workflow (`Build / SparkR-only`, 
windows-2025) intermittently hangs in the **"Run SparkR tests"** step while the 
R process waits on the SparkR backend socket handshake during session 
initialization. With no step-level timeout, a stalled handshake silently 
consumes the full 120-minute job timeout and surfaces as a *cancellation* 
rather than a fast, retryable failure — this alternates roughly every other 
scheduled run.
    
    This adds `timeout-minutes: 60` to the "Run SparkR tests" step so a stalled 
handshake fails fast; successful runs finish the step well under the cap.
    
    ### Why are the changes needed?
    
    Convert a silent 2-hour job cancel into a fast, retryable failure, and stop 
the workflow from alternating red/green due to a startup hang rather than a 
real regression.
    
    ### Does this PR introduce _any_ user-facing change?
    
    No. CI-workflow-only change.
    
    ### How was this patch tested?
    
    Fork CI (`build_sparkr_window` on the fork): ✅ **PASS** — 
https://github.com/HyukjinKwon/spark/actions/runs/28906197544
    Finished in ~1h20m with tests running to completion (`══ DONE ══`); the 
watchdog did not fire, confirming it is non-disruptive to healthy runs.
    
    JIRA: https://issues.apache.org/jira/browse/SPARK-58026
    
    Closes #57107 from HyukjinKwon/DO-NOT-MERGE/deflake-sparkr.
    
    Authored-by: Hyukjin Kwon <[email protected]>
    Signed-off-by: Hyukjin Kwon <[email protected]>
    (cherry picked from commit e0838ae41073fde3a998b85d1c80e38d418f586c)
    Signed-off-by: Hyukjin Kwon <[email protected]>
---
 .github/workflows/build_sparkr_window.yml | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/.github/workflows/build_sparkr_window.yml 
b/.github/workflows/build_sparkr_window.yml
index 8bbcdf79bd58..edd56a7e7d6c 100644
--- a/.github/workflows/build_sparkr_window.yml
+++ b/.github/workflows/build_sparkr_window.yml
@@ -79,6 +79,13 @@ jobs:
         mvn -DskipTests -Psparkr -Djna.nosys=true package -Dmaven.version=3.8.7
       shell: cmd
     - name: Run SparkR tests
+      # The SparkR test launch occasionally hangs while the R process waits on 
the
+      # SparkR backend socket handshake on the windows-2025 runner. Without a
+      # step-level bound this silently consumes the full 120-minute job 
timeout and
+      # is reported as a cancellation rather than a fast, retryable failure. 
Cap the
+      # step so a stalled handshake fails fast; successful runs finish it in 
well
+      # under 60 minutes (the whole job normally completes in ~75-85 minutes).
+      timeout-minutes: 60
       run: |
         set HADOOP_HOME=%USERPROFILE%\hadoop-3.3.6
         set PATH=%HADOOP_HOME%\bin;%PATH%


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to