Parquet file in invalid state while writing to S3 from EMR

I'm writing parquet to S3 from Spark 1.6.1 on EMR. And when it got to the last few files to write to S3, I received this stacktrace in the log with no other errors before or after it. It's very consistent. This particular batch keeps erroring the same way.

```
2016-06-10 01:46:05,282] WARN org.apache.spark.scheduler.TaskSetManager [task-result-getter-2hread] - Lost task 3737.0 in stage 2.0 (TID 10585, ip-172-16-96-32.ec2.internal): org.apache.spark.SparkException: Task failed while writing rows.
at org.apache.spark.sql.execution.datasources.DynamicPartitionWriterContainer.writeRows(WriterContainer.scala:414)
at org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelation$$anonfun$run$1$$anonfun$apply$mcV$sp$3.apply(InsertIntoHadoopFsRelation.scala:150)
at org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelation$$anonfun$run$1$$anonfun$apply$mcV$sp$3.apply(InsertIntoHadoopFsRelation.scala:150)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
at org.apache.spark.scheduler.Task.run(Task.scala:89)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.IOException: The file being written is in an invalid state. Probably caused by an error thrown previously. Current state: COLUMN
at org.apache.parquet.hadoop.ParquetFileWriter$STATE.error(ParquetFileWriter.java:146)
at org.apache.parquet.hadoop.ParquetFileWriter$STATE.startBlock(ParquetFileWriter.java:138)
at org.apache.parquet.hadoop.ParquetFileWriter.startBlock(ParquetFileWriter.java:195)
at org.apache.parquet.hadoop.InternalParquetRecordWriter.flushRowGroupToStore(InternalParquetRecordWriter.java:153)
at org.apache.parquet.hadoop.InternalParquetRecordWriter.close(InternalParquetRecordWriter.java:113)
at org.apache.parquet.hadoop.ParquetRecordWriter.close(ParquetRecordWriter.java:112)
at org.apache.spark.sql.execution.datasources.parquet.ParquetOutputWriter.close(ParquetRelation.scala:101)
at org.apache.spark.sql.execution.datasources.DynamicPartitionWriterContainer.writeRows(WriterContainer.scala:405)
... 8 more
```

**Reporter**: [Peter Halliday](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=hoangelos)

_{**Note**: *This issue was originally created as [PARQUET-632](https://issues.apache.org/jira/browse/PARQUET-632). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

从 stack trace 中提到的 ParquetFileWriter.java、InternalParquetRecordWriter.java 和 Spark 的 WriterContainer.scala 路径开始。复现 Spark 1.6.1-on-EMR 将 Parquet 文件写入 S3 的过程，并追踪导致 writer 停留在 COLUMN 状态的更早失败；完成的标准是确定一个可复现的原因，以及经过验证的修复或准确的诊断。

由索引模型根据 Issue 内容生成。