Github user dibbhatt commented on a diff in the pull request:
https://github.com/apache/spark/pull/6990#discussion_r33763939
--- Diff: core/src/main/scala/org/apache/spark/storage/BlockManager.scala
---
@@ -833,8 +833,10 @@ private[spark] class BlockManager(
logDebug("Put block %s locally took %s".format(blockId,
Utils.getUsedTimeMs(startTimeMs)))
// Either we're storing bytes and we asynchronously started
replication, or we're storing
- // values and need to serialize and replicate them now:
- if (putLevel.replication > 1) {
+ // values and need to serialize and replicate them now.
+ // Should not replicate the block if its StorageLevel is
StorageLevel.NONE or
+ // putting it to local is failed.
+ if (!putBlockInfo.isFailed && putLevel.replication > 1) {
--- End diff --
Also as I see the code path of doing rdd.Cache , the CacheManager first
check if block can be unrolled safely , and then only it tries to call
blockManager.putArray(..) which returns the updatedBlocks object. If block is
not unrolled safely , CacheManager wont cache the block. So I do not think PR
will de-optimize existing code path , as in RDD partitions case, once block is
unrolled safely , CacheManager store the block and get the updatedBlocks back,
and in this case my fix wont impact anything and replication will happen if
block is unrolled safely. @squito do let me know if this is correct.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
