Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/19041#discussion_r178964472
--- Diff:
core/src/main/scala/org/apache/spark/storage/BlockManagerMasterEndpoint.scala
---
@@ -250,6 +255,44 @@ class BlockManagerMasterEndpoint(
blockManagerIdByExecutor.get(execId).foreach(removeBlockManager)
}
+ private def recoverLatestRDDBlock(
+ execId: String,
+ excludeExecutors: Seq[String],
+ context: RpcCallContext): Unit = {
+ logDebug(s"Replicating first cached block on $execId")
+ val excluded = excludeExecutors.flatMap(blockManagerIdByExecutor.get)
+ val response: Option[Future[Boolean]] = for {
+ blockManagerId <- blockManagerIdByExecutor.get(execId)
+ info <- blockManagerInfo.get(blockManagerId)
+ blocks = info.cachedBlocks.collect { case r: RDDBlockId => r }
+ // As a heuristic, prioritize replicating the latest rdd. If this
succeeds,
+ // CacheRecoveryManager will try to replicate the remaining rdds.
--- End diff --
rather than the latest rdd, it would actually make more sense to take
advantage of the LRU already in the MemoryStore:
https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/storage/memory/MemoryStore.scala#L91
but maybe that is not easy to expose.
But I think that also means that the *receiving* end will put the
replicated block at the back of that LinkedHashMap, even though it really
hasn't been accessed at all.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
