You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2022/05/20 08:21:13 UTC
[GitHub] [spark] Ngone51 commented on a diff in pull request #36512: [SPARK-39152][CORE] Deregistering disk persisted local RDD blocks in case of IO related errors
Ngone51 commented on code in PR #36512:
URL: https://github.com/apache/spark/pull/36512#discussion_r877873226
##########
core/src/main/scala/org/apache/spark/storage/BlockManager.scala:
##########
@@ -933,46 +935,56 @@ private[spark] class BlockManager(
})
Some(new BlockResult(ci, DataReadMethod.Memory, info.size))
} else if (level.useDisk && diskStore.contains(blockId)) {
+ var diskData: BlockData = null
try {
- val diskData = diskStore.getBytes(blockId)
- val iterToReturn: Iterator[Any] = {
- if (level.deserialized) {
- val diskValues = serializerManager.dataDeserializeStream(
- blockId,
- diskData.toInputStream())(info.classTag)
- maybeCacheDiskValuesInMemory(info, blockId, level, diskValues)
- } else {
- val stream = maybeCacheDiskBytesInMemory(info, blockId, level, diskData)
- .map { _.toInputStream(dispose = false) }
- .getOrElse { diskData.toInputStream() }
- serializerManager.dataDeserializeStream(blockId, stream)(info.classTag)
- }
+ diskData = diskStore.getBytes(blockId)
+ val iterToReturn = if (level.deserialized) {
+ val diskValues = serializerManager.dataDeserializeStream(
+ blockId,
+ diskData.toInputStream())(info.classTag)
+ maybeCacheDiskValuesInMemory(info, blockId, level, diskValues)
+ } else {
+ val stream = maybeCacheDiskBytesInMemory(info, blockId, level, diskData)
+ .map { _.toInputStream(dispose = false) }
+ .getOrElse { diskData.toInputStream() }
+ serializerManager.dataDeserializeStream(blockId, stream)(info.classTag)
}
val ci = CompletionIterator[Any, Iterator[Any]](iterToReturn, {
releaseLockAndDispose(blockId, diskData, taskContext)
})
Some(new BlockResult(ci, DataReadMethod.Disk, info.size))
} catch {
- case ex: KryoException if ex.getCause.isInstanceOf[IOException] =>
- // We need to have detailed log message to catch environmental problems easily.
- // Further details: https://issues.apache.org/jira/browse/SPARK-37710
- processKryoException(ex, blockId)
- throw ex
+ case t: Throwable =>
+ if (diskData != null) {
+ diskData.dispose()
+ diskData = null
+ }
+ releaseLock(blockId, taskContext)
+ if (isIORelatedException(t)) {
+ logInfo(extendMessageWithBlockDetails(t.getMessage, blockId))
+ // Remove the block so that its unavailability is reported to the driver
+ removeBlock(blockId)
Review Comment:
This not only affects RDD blocks?
##########
core/src/main/scala/org/apache/spark/storage/BlockManager.scala:
##########
@@ -933,10 +933,29 @@ private[spark] class BlockManager(
})
Some(new BlockResult(ci, DataReadMethod.Memory, info.size))
} else if (level.useDisk && diskStore.contains(blockId)) {
- try {
- val diskData = diskStore.getBytes(blockId)
- val iterToReturn: Iterator[Any] = {
- if (level.deserialized) {
+ var retryCount = 0
+ val retryLimit = 3
Review Comment:
+1 to not retry.
--
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-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org