Github user tdas commented on a diff in the pull request: https://github.com/apache/spark/pull/17216#discussion_r106059434 --- Diff: sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamSuite.scala --- @@ -389,6 +392,102 @@ class StreamSuite extends StreamTest { query.stop() assert(query.exception.isEmpty) } + + test("SPARK-19873: streaming aggregation with change in number of partitions") { + val inputData = MemoryStream[(Int, Int)] + val agg = inputData.toDS().groupBy("_1").count() + + testStream(agg, OutputMode.Complete())( + AddData(inputData, (1, 0), (2, 0)), + StartStream(additionalConfs = Map(SQLConf.SHUFFLE_PARTITIONS.key -> "2")), + CheckAnswer((1, 1), (2, 1)), + StopStream, + AddData(inputData, (3, 0), (2, 0)), + StartStream(additionalConfs = Map(SQLConf.SHUFFLE_PARTITIONS.key -> "5")), + CheckAnswer((1, 1), (2, 2), (3, 1)), + StopStream, + AddData(inputData, (3, 0), (1, 0)), + StartStream(additionalConfs = Map(SQLConf.SHUFFLE_PARTITIONS.key -> "1")), + CheckAnswer((1, 2), (2, 2), (3, 2))) + } + + test("SPARK-19873: backward compatibility - recover from a Spark v2.1 checkpoint") { + var inputData: MemoryStream[Int] = null + var query: DataStreamWriter[Row] = null + + def init(): Unit = { + inputData = MemoryStream[Int] + inputData.addData(1, 2, 3, 4) + inputData.addData(3, 4, 5, 6) + inputData.addData(5, 6, 7, 8) + + query = inputData + .toDF() + .groupBy($"value") + .agg(count("*")) + .writeStream + .outputMode("complete") + .format("memory") + } + + // Get an existing checkpoint generated by Spark v2.1. + // v2.1 does not record # shuffle partitions in the offset metadata. + val resourceUri = + this.getClass.getResource("/structured-streaming/checkpoint-version-2.1.0").toURI + val checkpointDir = new File(resourceUri) + + // 1 - Test if recovery from the checkpoint is successful. + init() + withTempDir(dir => { + // Copy the checkpoint to a temp dir to prevent changes to the original. + // Not doing this will lead to the test passing on the first run, but fail subsequent runs. + FileUtils.copyDirectory(checkpointDir, dir) + + // Checkpoint data was generated by a query with 10 shuffle partitions. + // In order to test reading from the checkpoint, the checkpoint must have two or more batches, + // since the last batch may be rerun. + withSQLConf(SQLConf.SHUFFLE_PARTITIONS.key -> "10") { + var streamingQuery: StreamingQuery = null + try { + streamingQuery = + query.queryName("counts").option("checkpointLocation", dir.getCanonicalPath).start() + streamingQuery.processAllAvailable() + inputData.addData(9) + streamingQuery.processAllAvailable() + + QueryTest.checkAnswer(spark.table("counts").toDF(), + Row("1", 1) :: Row("2", 1) :: Row("3", 2) :: Row("4", 2) :: + Row("5", 2) :: Row("6", 2) :: Row("7", 1) :: Row("8", 1) :: Row("9", 1) :: Nil) + } finally { + if (streamingQuery ne null) { + streamingQuery.stop() + } + } + } + }) + + // 2 - Check recovery with wrong num shuffle partitions + init() + withTempDir(dir => { + FileUtils.copyDirectory(checkpointDir, dir) + + // Since the number of partitions is greater than 10, should throw exception. + withSQLConf(SQLConf.SHUFFLE_PARTITIONS.key -> "15") { + var streamingQuery: StreamingQuery = null + try { + intercept[StreamingQueryException] { --- End diff -- what is the error message?
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org