spark-rapids icon indicating copy to clipboard operation
spark-rapids copied to clipboard

Add spark350emr shim layer [EMR]

Open mimaomao opened this issue 1 year ago • 3 comments

This PR targets to add a new shim layer spark350emr which supports running Spark RAPIDS on AWS EMR Spark 3.5.0.


Note, this PR is a new revision of previous PR rebased on branch-24.04. You can find more details about testing in that PR.

mimaomao avatar Feb 22 '24 04:02 mimaomao

Somewhere in ./jenkins/version-def.sh or in .github/workflows/mvn-verify-check.yml we need to exclude 350emr from package-tests matrix on GH hosted PR checks because we won't have access to 3.5.0-amzn-0 dependencies presumably.

Error:  Failed to execute goal on project rapids-4-spark-emr-bom: Could not resolve dependencies for project com.nvidia:rapids-4-spark-emr-bom:pom:24.04.0-SNAPSHOT: The following artifacts could not be resolved: org.apache.spark:spark-sql_2.12:jar:3.5.0-amzn-0, org.apache.spark:spark-hive_2.12:jar:3.5.0-amzn-0: org.apache.spark:spark-sql_2.12:jar:3.5.0-amzn-0 was not found in https://repo1.maven.org/maven2 during a previous attempt. This failure was cached in the local repository and resolution is not reattempted until the update interval of central has elapsed or updates are forced -> [Help 1]

We will also need to produce a spark-rapids-private shim for 350emr

cc @GaryShen2008 @sameerz

gerashegalov avatar Feb 23 '24 18:02 gerashegalov

Somewhere in ./jenkins/version-def.sh or in .github/workflows/mvn-verify-check.yml we need to exclude 350emr from package-tests matrix on GH hosted PR checks because we won't have access to 3.5.0-amzn-0 dependencies presumably.

Error:  Failed to execute goal on project rapids-4-spark-emr-bom: Could not resolve dependencies for project com.nvidia:rapids-4-spark-emr-bom:pom:24.04.0-SNAPSHOT: The following artifacts could not be resolved: org.apache.spark:spark-sql_2.12:jar:3.5.0-amzn-0, org.apache.spark:spark-hive_2.12:jar:3.5.0-amzn-0: org.apache.spark:spark-sql_2.12:jar:3.5.0-amzn-0 was not found in https://repo1.maven.org/maven2 during a previous attempt. This failure was cached in the local repository and resolution is not reattempted until the update interval of central has elapsed or updates are forced -> [Help 1]

We will also need to produce a spark-rapids-private shim for 350emr

cc @GaryShen2008 @sameerz

EMR pre-merge/nightly build&test will be like what we've done for Databricks runtime.

We'll need separated CI jobs running on EMR

NvTimLiu avatar Mar 12 '24 00:03 NvTimLiu

Retarget to branch-24.06 for next release, as we're running v24.04 release, please let me know if you've any concern, thanks!

NvTimLiu avatar Apr 15 '24 03:04 NvTimLiu