+1 (binding) verified signature, checksum, license. The checksum was generated using the old buggy release script because it was executed in the 0.13.x branch so it still used the full file path. I have updated it to use the relative file path. In case anyone sees checksum failure, please re-download the checksum file and verify again.
Ran unit tests for all engine versions and JDK versions, AWS Integration tests. For the Spark flaky test, given #4033 fixes the issue and it was not a bug of the source code, I think we can continue without re-cut a candidate. Tested basic operations, copy-on-write delete, update and rewrite data files on AWS EMR Spark 3.1 Flink 1.14 and verified fixes #3986 and #4024. I did some basic tests for #4023 (the predicate pushdown fix) but I don't have a large Spark 3.2 installation to further verify the performance. It would be great if anyone else could do some additional verifications. Best, Jack Ye On Fri, Feb 11, 2022 at 8:24 PM Manong Karl <abc549...@gmail.com> wrote: > It's flaky. This exception is only found in one agent of TeamCity. > Changing agents will resolve the issue. > > Ryan Blue <b...@tabular.io> 于2022年2月12日周六 08:57写道： > >> Does that exception fail consistently, or is it a flaky test? We recently >> fixed another Spark test that was flaky because of sampling and sort order: >> https://github.com/apache/iceberg/pull/4033 >> >> On Thu, Feb 10, 2022 at 7:12 PM Manong Karl <abc549...@gmail.com> wrote: >> >>> I got an issue failed on spark 3.2 >>> TestMergeOnReadDelete.testDeleteWithSerializableIsolation[catalogName = >>> testhive, implementation = org.apache.iceberg.spark.SparkCatalog, config = >>> {type=hive, default-namespace=default}, format = orc, vectorized = true, >>> distributionMode = none] · Issue #4090 · apache/iceberg (github.com) >>> <https://github.com/apache/iceberg/issues/4090>. >>> Is it just my exception? >>> >> >> >> -- >> Ryan Blue >> Tabular >> >
