Github user tdas commented on a diff in the pull request:
https://github.com/apache/spark/pull/2366#discussion_r17921361
--- Diff: core/src/main/scala/org/apache/spark/storage/BlockManager.scala
---
@@ -787,31 +789,88 @@ private[spark] class BlockManager(
}
/**
+ * Get peer block managers in the system.
+ */
+ private def getPeers(forceFetch: Boolean): HashSet[BlockManagerId] = {
+ val cachedPeersTtl = conf.getInt("spark.storage.cachedPeersTtl", 60 *
1000) // milliseconds
+ val timeout = System.currentTimeMillis - lastPeerFetchTime >
cachedPeersTtl
+
+ cachedPeers.synchronized {
+ if (cachedPeers.isEmpty || forceFetch || timeout) {
+ cachedPeers.clear()
+ cachedPeers ++= master.getPeers(blockManagerId).sortBy(_.hashCode)
+ lastPeerFetchTime = System.currentTimeMillis
+ logDebug("Fetched peers from master: " + cachedPeers.mkString("[",
",", "]"))
+ }
+ }
+ cachedPeers
+ }
+
+ /**
* Replicate block to another node.
*/
- @volatile var cachedPeers: Seq[BlockManagerId] = null
private def replicate(blockId: BlockId, data: ByteBuffer, level:
StorageLevel): Unit = {
+ val maxReplicationFailures =
conf.getInt("spark.storage.maxReplicationFailures", 1)
+ val numPeersToReplicateTo = level.replication - 1
+ val peersReplicatedTo = new HashSet[BlockManagerId]
+ val peersFailedToReplicateTo = new HashSet[BlockManagerId]
val tLevel = StorageLevel(
level.useDisk, level.useMemory, level.useOffHeap,
level.deserialized, 1)
- if (cachedPeers == null) {
- cachedPeers = master.getPeers(blockManagerId, level.replication - 1)
+ val startTime = System.nanoTime
+ val random = new Random(blockId.hashCode)
+
+ var forceFetchPeers = false
+ var failures = 0
+ var done = false
+
+ // Get a random peer
+ def getRandomPeer(): Option[BlockManagerId] = {
+ val peers = getPeers(forceFetchPeers) -- peersReplicatedTo --
peersFailedToReplicateTo
+ if (!peers.isEmpty) Some(peers.toSeq(random.nextInt(peers.size)))
else None
}
- for (peer: BlockManagerId <- cachedPeers) {
- val start = System.nanoTime
- data.rewind()
- logDebug(s"Try to replicate $blockId once; The size of the data is
${data.limit()} Bytes. " +
- s"To node: $peer")
- try {
- blockTransferService.uploadBlockSync(
- peer.host, peer.port, blockId.toString, new
NioByteBufferManagedBuffer(data), tLevel)
- } catch {
- case e: Exception =>
- logError(s"Failed to replicate block to $peer", e)
+ // One by one choose a random peer and try uploading the block to it
+ // If replication fails (e.g., target peer is down), force the list of
cached peers
+ // to be re-fetched from driver and then pick another random peer for
replication. Also
+ // temporarily black list the peer for which replication failed.
+ while (!done) {
+ getRandomPeer() match {
+ case Some(peer) =>
+ try {
+ val onePeerStartTime = System.nanoTime
+ data.rewind()
+ logTrace(s"Trying to replicate $blockId of ${data.limit()}
bytes to $peer")
+ blockTransferService.uploadBlockSync(
+ peer.host, peer.port, blockId.toString, new
NioByteBufferManagedBuffer(data), tLevel)
+ logTrace(s"Replicated $blockId of ${data.limit()} bytes to
$peer in %f ms"
+ .format((System.nanoTime - onePeerStartTime) / 1e6))
+ peersReplicatedTo += peer
+ forceFetchPeers = false
+ if (peersReplicatedTo.size == numPeersToReplicateTo) {
+ done = true
+ }
+ } catch {
+ case e: Exception =>
+ logWarning(s"Failed to replicate $blockId to $peer, failure
#$failures", e)
+ failures += 1
+ forceFetchPeers = true
+ peersFailedToReplicateTo += peer
+ if (failures > maxReplicationFailures) {
+ done = true
+ }
+ }
+ case None =>
+ // no peer left to replicate to
+ done = true
--- End diff --
The initial list wont ever have self as the BlockManagerMaster returns list
of nodes excluding the id of the node requesting it (that is, self).
Nonetheless getPeers can return empty list (e.g., local mode, with only one
BlockManager). And can also happen in the bootstrapping time. However, current
Spark already suffers from this problem. In fact its much worse. Currently, the
peer list is only fetched once from master ( upon first time replication) is
never updated every again! So this patch is a strict improvement, as the peer
list updated, by default, every minute.
---
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
