[
https://issues.apache.org/jira/browse/SPARK-36766?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17426152#comment-17426152
]
Steve Loughran commented on SPARK-36766:
----------------------------------------
I can see why you'd want to do this (consistent URLs on EMR and ASF spark
builds), but be aware that there's a risk the s3a connector might assume its
got an s3a URL, and sometimes it'll leak.
> Spark SQL DDL does not recognize fs.s3.impl implied filesystem in LOCATION tag
> ------------------------------------------------------------------------------
>
> Key: SPARK-36766
> URL: https://issues.apache.org/jira/browse/SPARK-36766
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 3.0.2
> Reporter: Sungwon
> Priority: Major
>
> CREATE TABLE with the LOCATION tag via SPARK SQL DDL does not seem to
> recognize the fileSystem implied by setting fs.s3.impl.
> The environment setting is set to:
> |spark.hadoop.fs.s3.impl|org.apache.hadoop.fs.s3a.S3AFileSystem|
> The following spark.sql ddl raises an exception:
> spark.sql("""CREATE TABLE local.db.table (
> col1 string,
> col2 string,
> )
> LOCATION 's3://bucket/prefix'
> """)
> Exception:
> Py4JJavaError: An error occurred while calling o86.sql. :
> org.apache.iceberg.exceptions.CommitStateUnknownException: Got exception:
> org.apache.hadoop.fs.UnsupportedFileSystemException No FileSystem for scheme
> "s3"
> This is inconsistent with other spark functions such as
> spark.read.parquet("s3://bucket/prefix") which succeeds with the same
> configuration
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
