[ 
https://issues.apache.org/jira/browse/SPARK-24375?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16568664#comment-16568664
 ] 

Mridul Muralidharan commented on SPARK-24375:
---------------------------------------------

{quote}
It's not desired behavior to catch exception thrown by TaskContext.barrier() 
silently. However, in case this really happens, we can detect that because we 
have `epoch` both in driver side and executor side, more details will go to the 
design doc of BarrierTaskContext.barrier() SPARK-24581
{quote}

The current 'barrier' function does not identify 'which' barrier it is from a 
user point of view.

Here, due to exceptions raised (not necessarily from barrier(), but could be 
from user code as well), different tasks are waiting on different barriers.

{code}
try {
  ... snippet A ...
  // Barrier 1
  context.barrier()
  ... snippet B ...
} catch { ... }
... snippet C ...
// Barrier 2
context.barrier()
{code}

T1 waits on barrier 1, T2 could have raised exception in snippet A and ends up 
waiting on Barrier 2 (having never seen Barrier 1).
In this scenario, how is spark making progress ?
(And ofcourse, when T1 reaches barrier 2, when T2 has moved past it).
I did not see this clarified in the design or in the implementation.


> Design sketch: support barrier scheduling in Apache Spark
> ---------------------------------------------------------
>
>                 Key: SPARK-24375
>                 URL: https://issues.apache.org/jira/browse/SPARK-24375
>             Project: Spark
>          Issue Type: Story
>          Components: Spark Core
>    Affects Versions: 3.0.0
>            Reporter: Xiangrui Meng
>            Assignee: Jiang Xingbo
>            Priority: Major
>
> This task is to outline a design sketch for the barrier scheduling SPIP 
> discussion. It doesn't need to be a complete design before the vote. But it 
> should at least cover both Scala/Java and PySpark.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

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

Reply via email to