Skip to content

Conversation

@oliviertoupin
Copy link

Tested with our workload with success.

Bring last version of parquet-mr to spark.

Might fix => SPARK-4258

@AmplabJenkins
Copy link

Can one of the admins verify this patch?

@JoshRosen
Copy link
Contributor

We just bumped to 1.7 in master (#6597); can you check whether that PR resolved SPARK-4258?

@kostya-sh
Copy link

Parquet 1.6.0 certainly fixes SPARK-4258, we've been using a custom build of Spark with parquet 1.6.0rc7 for a while now. We had to apply similar fix to the one from this pull request. See https://issues.apache.org/jira/browse/SPARK-6566

If parquet 1.7 is not coming to Spark 1.4.x it would be really good to at least update it to parquet 1.6.0

@rxin
Copy link
Contributor

rxin commented Jun 5, 2015

hey @kostya-sh - in the past parquet minor bumps have introduced bugs, and as a result we are very conservative w.r.t. upgrading Parquet. However, I think you can always specify a higher version of the Parquet jar in your user program, and then set the classloader to prefer user jars first. Alternatively, you can also just bump the Parquet version in your custom build of Spark. Does that make sense?

@kostya-sh
Copy link

Unfortunately a small change to Spark source code is required when updating from 1.6.0rc3 to 1.6.0, so it is not possible to just update parquet version in the user program. We bumped Parquet version in our custom build of Spark (and fixed the compilation issue I mentioned) instead.

@oliviertoupin
Copy link
Author

Closing because fixed by #6597

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants