[ https://issues.apache.org/jira/browse/BEAM-8384?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16949333#comment-16949333 ]
Ryan Skraba commented on BEAM-8384: ----------------------------------- Related to BEAM-8191 (exploding number of partitions during flatten). There could definitely be some cleanup around [getPartitioner|https://github.com/apache/beam/blob/a5e7e671d4571d86991151e79586c98fd107a2b1/runners/spark/src/main/java/org/apache/beam/runners/spark/translation/TransformTranslator.java#L571] and how bundleSize impacts the partitioner choice. > Spark runner is not respecting spark.default.parallelism user defined > configuration > ----------------------------------------------------------------------------------- > > Key: BEAM-8384 > URL: https://issues.apache.org/jira/browse/BEAM-8384 > Project: Beam > Issue Type: Bug > Components: runner-spark > Affects Versions: 2.16.0 > Reporter: Ismaël Mejía > Assignee: Ismaël Mejía > Priority: Major > Fix For: 2.17.0 > > > It was reported in [the mailing > list|https://lists.apache.org/thread.html/792fb7fc2a5113837fbcdafce6a5d9100309881b366c1a7163d2c898@%3Cdev.beam.apache.org%3E] > that the Spark runner is not respecting the user defined Spark default > parallelism configuration. We should investigate and if it is the case ensure > that a user defined configuration is always respected. Runner optimizations > should apply only for default (unconfigured) values otherwise we will confuse > users and limit them from parametrizing Spark for their best convenience. -- This message was sent by Atlassian Jira (v8.3.4#803005)