[
https://issues.apache.org/jira/browse/SPARK-9896?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15009920#comment-15009920
]
Mark Grover commented on SPARK-9896:
------------------------------------
Posting here in case someone runs into this as well. Based on my experience,
this happens when the client configuration for accessing HDFS/S3 is incorrect
(which was the case for me) or if possibly if HDFS is inaccessible.
In my case, I was accessing a remote non-secure Hadoop cluster from a node
which had secure HDFS configuration. So, while the error message should be
something more relevant and unrelated to Parquet, this inaccessibility shows up
as a Parquet metadata error. For me, I fixed it by doing a kdestroy on the
gateway/client node and was able to run it fine.
If you are hitting this, try accessing s3 or hdfs from that node, using the
client configuration (say using 'hadoop fs -ls <path>'). If that doesn't
succeed, that's your root cause.
I will file a separate JIRA for improving the error message.
> Parquet Schema Assertion
> ------------------------
>
> Key: SPARK-9896
> URL: https://issues.apache.org/jira/browse/SPARK-9896
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Reporter: Michael Armbrust
> Priority: Blocker
>
> Need to investigate more, but I'm seeing this all of a sudden.
> {code}
> java.lang.AssertionError: assertion failed: No predefined schema found, and
> no Parquet data files or summary files found under
> s3n:/.../databricks-performance-datasets/tpcds/sf1500-parquet/useDecimal=true/parquet/item
> {code}
> Possibly related to [SPARK-9407].
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
