You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2019/03/16 04:10:29 UTC

[GitHub] [spark] felixcheung commented on a change in pull request #24017: [SPARK-27094][YARN] Work around RackResolver swallowing thread interrupt.

felixcheung commented on a change in pull request #24017: [SPARK-27094][YARN] Work around RackResolver swallowing thread interrupt.
URL: https://github.com/apache/spark/pull/24017#discussion_r266188431
 
 

 ##########
 File path: resource-managers/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala
 ##########
 @@ -417,12 +424,40 @@ private[yarn] class YarnAllocator(
         containersToUse, remainingAfterHostMatches)
     }
 
-    // Match remaining by rack
+    // Match remaining by rack. Because YARN's RackResolver swallows thread interrupts
+    // (see SPARK-27094), which can cause this code to miss interrupts from the AM, use
+    // a separate thread to perform the operation.
     val remainingAfterRackMatches = new ArrayBuffer[Container]
-    for (allocatedContainer <- remainingAfterHostMatches) {
-      val rack = resolver.resolve(conf, allocatedContainer.getNodeId.getHost)
-      matchContainerToRequest(allocatedContainer, rack, containersToUse,
-        remainingAfterRackMatches)
+    if (remainingAfterHostMatches.nonEmpty) {
+      var exception: Option[Throwable] = None
+      val thread = new Thread("spark-rack-resolver") {
+        override def run(): Unit = {
+          try {
+            for (allocatedContainer <- remainingAfterHostMatches) {
+              val rack = resolver.resolve(conf, allocatedContainer.getNodeId.getHost)
+              matchContainerToRequest(allocatedContainer, rack, containersToUse,
+                remainingAfterRackMatches)
+            }
+          } catch {
+            case NonFatal(e) =>
+              exception = Some(e)
 
 Review comment:
   should it capture all exception/error when it is fatal (if possible)
   I can see how the error can be "lost" on this new thread instead of getting propagate to re-throw from the original (the one calling thread.join())

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org