HeartSaVioR commented on code in PR #50559:
URL: https://github.com/apache/spark/pull/50559#discussion_r2043006686
##########
sql/core/src/test/scala/org/apache/spark/sql/execution/streaming/state/RocksDBCheckpointFailureInjectionSuite.scala:
##########
@@ -414,6 +414,84 @@ class RocksDBCheckpointFailureInjectionSuite extends
StreamTest
}
}
+ case class FailureConf2(logType: String, checkpointFormatVersion: String) {
+ override def toString: String = {
+ s"logType = $logType, checkpointFormatVersion = $checkpointFormatVersion"
+ }
+ }
+
+ // tests to validate the behavior after failures when writing to the commit
and offset logs
+ Seq(
+ FailureConf2("commits", checkpointFormatVersion = "1"),
+ FailureConf2("commits", checkpointFormatVersion = "2"),
+ FailureConf2("offsets", checkpointFormatVersion = "1"),
+ FailureConf2("offsets", checkpointFormatVersion = "2")).foreach {
failureConf =>
+ test(s"Progress log fails to write $failureConf") {
+ val hadoopConf = new Configuration()
+ hadoopConf.set(STREAMING_CHECKPOINT_FILE_MANAGER_CLASS.parent.key,
fileManagerClassName)
+ val rocksdbChangelogCheckpointingConfKey =
+ RocksDBConf.ROCKSDB_SQL_CONF_NAME_PREFIX +
".changelogCheckpointing.enabled"
+
+ withTempDirAllowFailureInjection { (checkpointDir, injectionState) =>
+ withSQLConf(
+ rocksdbChangelogCheckpointingConfKey -> "true",
+ SQLConf.STATE_STORE_MIN_DELTAS_FOR_SNAPSHOT.key -> "2") {
+ val inputData = MemoryStream[Int]
+ val aggregated =
+ inputData.toDF()
+ .groupBy($"value")
+ .agg(count("*"))
+ .as[(Int, Long)]
+
+ // This should cause the second batch to fail
+ injectionState.createAtomicDelayCloseRegex =
Seq(s".*/${failureConf.logType}/1")
+
+ val additionalConfs = Map(
+ rocksdbChangelogCheckpointingConfKey -> "true",
+ SQLConf.STATE_STORE_CHECKPOINT_FORMAT_VERSION.key ->
+ failureConf.checkpointFormatVersion,
+ STREAMING_CHECKPOINT_FILE_MANAGER_CLASS.parent.key ->
fileManagerClassName)
+
+ testStream(aggregated, Update)(
+ StartStream(
+ checkpointLocation = checkpointDir.getAbsolutePath,
+ additionalConfs = additionalConfs),
+ AddData(inputData, 3),
+ CheckLastBatch((3, 1)),
+ AddData(inputData, 3, 2),
+ // We should categorize this error.
+ // TODO after the error is categorized, we should check error class
+ ExpectFailure[IOException] { _ => () }
+ )
+
+ injectionState.createAtomicDelayCloseRegex = Seq.empty
+
+ inputData.addData(3, 1)
+
+ // The query will restart successfully and start at the checkpoint
after Batch 1
+ testStream(aggregated, Update)(
+ StartStream(
+ checkpointLocation = checkpointDir.getAbsolutePath,
+ additionalConfs = additionalConfs),
+ AddData(inputData, 4),
+ if (failureConf.logType == "commits") {
+ // If the failure is in the commit log, data is already
committed. The batch will
Review Comment:
This is very uneasy to follow, because the behavior of MemoryStream seems to
impact a lot on the test. It's not easy to reason about when the commit on the
source happens and how MemoryStream will behave. It'd be harder than the origin
test logic.
I'd test with file stream where we only append files and Spark is expected
to process all files "regardless of these failures". Once we change the output
mode to complete, we should see the same result in the latest batch which
processes all files.
Also, I'm not comfortable with the behavior - `If the failure is in the
commit log, data is already committed.`. Shouldn't we commit the offset for
batch N to source when batch N is NOT committed? I suspect this is an
indication of data loss/correctness - I wish I'm misunderstanding something.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
