Github user tdas commented on a diff in the pull request:
https://github.com/apache/spark/pull/2366#discussion_r17498158
--- Diff:
core/src/main/scala/org/apache/spark/storage/BlockManagerMasterActor.scala ---
@@ -403,16 +402,20 @@ class BlockManagerMasterActor(val isLocal: Boolean,
conf: SparkConf, listenerBus
blockIds.map(blockId => getLocations(blockId))
}
- private def getPeers(blockManagerId: BlockManagerId, size: Int):
Seq[BlockManagerId] = {
- val peers: Array[BlockManagerId] = blockManagerInfo.keySet.toArray
-
- val selfIndex = peers.indexOf(blockManagerId)
+ /** Get the list of the peers of the given block manager */
+ private def getPeers(blockManagerId: BlockManagerId):
Seq[BlockManagerId] = {
+ val blockManagerIds = blockManagerInfo.keySet.filterNot { _.isDriver
}.toArray
+ val selfIndex = blockManagerIds.indexOf(blockManagerId)
if (selfIndex == -1) {
- throw new SparkException("Self index for " + blockManagerId + " not
found")
+ logError("Self index for " + blockManagerId + " not found")
+ Seq.empty
+ } else {
+ // If the blockManagerIds is [ id1 id2 id3 id4 id5 ] and the
blockManagerId is id2
+ // Then this code will return the list [ id3 id4 id5 id1 ]
+ Array.tabulate[BlockManagerId](blockManagerIds.size - 1) { i =>
+ blockManagerIds((selfIndex + i + 1) % blockManagerIds.size)
--- End diff --
Again, I had just preserved the existing logic here. However, since I am
changing the overall logic from deterministic peer selection (where I want to
maintain the order they are returned, so that no two nodes use the same 3rd
node as replication target) to random peer selection, I think your suggestion
makes sense.
---
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
