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

pan3793 pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/spark.git


The following commit(s) were added to refs/heads/master by this push:
     new 07e77e6e17dd [SPARK-60612][CORE] Remove redundant dump command from 
async profiler
07e77e6e17dd is described below

commit 07e77e6e17dd67ae55e8792c418238bd0b25d0bf
Author: Zhen Wang <[email protected]>
AuthorDate: Fri Oct 30 16:04:40 2026 +0800

    [SPARK-60612][CORE] Remove redundant dump command from async profiler
    
    ### What changes were proposed in this pull request?
    
    Remove the redundant `dump` command and its definition from 
`SparkAsyncProfiler`, and update the related comment.
    
    ### Why are the changes needed?
    
    `SparkAsyncProfiler.writeChunk` executes `dump` immediately after `stop`, 
but this call has no effect for JFR output.
    
    In async-profiler:
    - 
[`Profiler::stop`](https://github.com/async-profiler/async-profiler/blob/v4.5/src/profiler.cpp#L1096)
 stops the JFR recording and sets `_state = IDLE`.
    - 
[`Profiler::dump`](https://github.com/async-profiler/async-profiler/blob/v4.5/src/profiler.cpp#L1144-L1150)
 only flushes JFR data when `_state == RUNNING`.
    
    Since the profiler is already in the `IDLE` state, the subsequent `dump` 
does nothing.
    
    ### Does this PR introduce _any_ user-facing change?
    
    No.
    
    ### How was this patch tested?
    
    Manually verified that profile files are successfully generated after 
applying this change and can be parsed into flame graphs.
    
    ### Was this patch authored or co-authored using generative AI tooling?
    
    Generated-by: OpenAI Codex (GPT-6)
    
    Closes #59858 from wForget/SPARK-60612.
    
    Authored-by: Zhen Wang <[email protected]>
    Signed-off-by: Cheng Pan <[email protected]>
---
 .../main/scala/org/apache/spark/profiler/SparkAsyncProfiler.scala   | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git 
a/connector/profiler/src/main/scala/org/apache/spark/profiler/SparkAsyncProfiler.scala
 
b/connector/profiler/src/main/scala/org/apache/spark/profiler/SparkAsyncProfiler.scala
index f86b47e98d8f..39ff33cea9ee 100644
--- 
a/connector/profiler/src/main/scala/org/apache/spark/profiler/SparkAsyncProfiler.scala
+++ 
b/connector/profiler/src/main/scala/org/apache/spark/profiler/SparkAsyncProfiler.scala
@@ -52,7 +52,6 @@ private[spark] class SparkAsyncProfiler(conf: SparkConf, 
executorId: String) ext
 
   private val startcmd = 
s"start,$profilerOptions,file=$profilerLocalDir/$profileFile"
   private val stopcmd = 
s"stop,$profilerOptions,file=$profilerLocalDir/$profileFile"
-  private val dumpcmd = 
s"dump,$profilerOptions,file=$profilerLocalDir/$profileFile"
   private val resumecmd = 
s"resume,$profilerOptions,file=$profilerLocalDir/$profileFile"
 
   private val PROFILER_FOLDER_PERMISSIONS = new 
FsPermission(Integer.parseInt("770", 8).toShort)
@@ -172,11 +171,10 @@ private[spark] class SparkAsyncProfiler(conf: SparkConf, 
executorId: String) ext
     }
 
     try {
-      // stop (pause) the profiler, dump the results and then resume. This is 
not ideal as we miss
-      // the events while the file is being dumped, but that is the only way 
to make sure that
+      // stop (pause) the profiler and then resume. This is not ideal as we 
miss
+      // the events while the profiler is stopped, but that is the only way to 
make sure that
       // the chunk of data we are copying to dfs is in a consistent state.
       profiler.get.execute(stopcmd)
-      profiler.get.execute(dumpcmd)
       var remaining = inputStream.available()
       if (!lastChunk) {
         profiler.get.execute(resumecmd)


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

Reply via email to