Github user gvramana commented on a diff in the pull request:
https://github.com/apache/carbondata/pull/1002#discussion_r120927111
--- Diff:
integration/spark-common/src/main/scala/org/apache/carbondata/spark/rdd/CarbonMergerRDD.scala
---
@@ -405,11 +411,16 @@ class CarbonMergerRDD[K, V](
NodeInfo(splitsPerNode.getTaskId,
splitsPerNode.getCarbonInputSplitList.size()))
if (blockletCount != 0) {
+ val taskInfo = splitInfo.asInstanceOf[CarbonInputSplitTaskInfo]
val multiBlockSplit = new
CarbonMultiBlockSplit(absoluteTableIdentifier,
-
splitInfo.asInstanceOf[CarbonInputSplitTaskInfo].getCarbonInputSplitList,
+ taskInfo.getCarbonInputSplitList,
Array(nodeName))
- result.add(new CarbonSparkPartition(id, partitionNo,
multiBlockSplit))
- partitionNo += 1
+ if (isPartitionTable) {
--- End diff --
This handling will not be sufficient,
When number of partitions(Example:100) is not equal to number of
nodes(Example:5) , getPartitions will divide total blocks among available
nodes. Then each node will get more than one taskno/partitionNo to handle.
Compute function in executor just merges all the given btrees(segid+taskid)
into one task. So multiple taskids/partitions will be merged to one. This
disturbs partition mapping.
---
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.
---
