[jira] [Commented] (SPARK-26851) CachedRDDBuilder only partially implements double-checked locking

["Bruce Robbins (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-26851?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16763818#comment-16763818 ] 

Bruce Robbins commented on SPARK-26851:
---------------------------------------

[~maropu] [~cloud_fan]

I will let this Jira marinate for a little while, since I might have missed something (misunderstanding of the code, improvements in the JMM since initial introduction). I will make a PR if needed, but I don't want to use up 4 hours of Jenkins resources for a nonsense PR.

> CachedRDDBuilder only partially implements double-checked locking
> -----------------------------------------------------------------
>
>                 Key: SPARK-26851
>                 URL: https://issues.apache.org/jira/browse/SPARK-26851
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 2.4.0, 3.0.0
>            Reporter: Bruce Robbins
>            Priority: Minor
>
> In CachedRDDBuilder, {{cachedColumnBuffers}} uses double-checked locking to lazily initialize {{_cachedColumnBuffers}}. Also, clearCache uses double-checked locking to likely avoid synchronization when {{_cachedColumnBuffers}} is still null.
> However, the resource (in this case, {{_cachedColumnBuffers}}) is not declared as volatile, which could cause some visibility problems in both methods. {{cachedColumnBuffers}} may see a null reference when there is actually a constructed RDD, and {{clearCache}} may pick up a reference to partially constructed RDD.
> From Java Concurrency in Practice by Brian Goetz et al:
> {quote}Subsequent changes in the JMM (Java 5.0 and later) have enabled DCL to work if resource is made volatile, and the performance impact of this is small since volatile reads are usually only slightly more expensive than nonvolatile reads.
> {quote}
> There are comments in other documentation that volatile is not needed if the resource in question (again, in this case {{_cachedColumnBuffers}}) is immutable. While an RDD is immutable from a Spark user's point of view, it may not be from a JVM's point of view, since not all internal fields are final.
> I've marked this as minor since the race conditions are highly unlikely.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org