Github user squito commented on the issue:

    https://github.com/apache/spark/pull/20998
  
    I'm not even really concerned about the case for two hosts -- I agree its 
fine if we do something sub-optimal.  I'm more concerned about code-clarity and 
the behavior in general.  It seems cleaner to me if speculation doesn't worry 
about where its failed before, and those exclusions are left to the blacklist.
    
    But it sounds like you're saying the prior behavior was really desirable -- 
you think its better if speculation always excludes hosts that task has ever 
failed on?  I'm happy to defer to your opinion on this, I haven't really 
stressed speculative execution yet.  Then lets just change that comment in the 
code to be consistent.


---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to