From 33e64ecacbc44567f9cba2644a30a118653ea5fa Mon Sep 17 00:00:00 2001
From: Rui Li <rui.li@intel.com>
Date: Wed, 16 Jul 2014 22:53:37 +0530
Subject: SPARK-2277: make TaskScheduler track hosts on rack

Hi mateiz, I've created [SPARK-2277](https://issues.apache.org/jira/browse/SPARK-2277) to make TaskScheduler track hosts on each rack. Please help to review, thanks.

Author: Rui Li <rui.li@intel.com>

Closes #1212 from lirui-intel/trackHostOnRack and squashes the following commits:

2b4bd0f [Rui Li] SPARK-2277: refine UT
fbde838 [Rui Li] SPARK-2277: add UT
7bbe658 [Rui Li] SPARK-2277: rename the method
5e4ef62 [Rui Li] SPARK-2277: remove unnecessary import
79ac750 [Rui Li] SPARK-2277: make TaskScheduler track hosts on rack
---
 .../org/apache/spark/scheduler/TaskSchedulerImpl.scala    | 15 +++++++++++++++
 .../scala/org/apache/spark/scheduler/TaskSetManager.scala | 10 +++++++---
 2 files changed, 22 insertions(+), 3 deletions(-)

(limited to 'core/src/main/scala')

diff --git a/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala b/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala
index 4b6d6da5a6..be3673c48e 100644
--- a/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala
+++ b/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala
@@ -88,6 +88,8 @@ private[spark] class TaskSchedulerImpl(
   // in turn is used to decide when we can attain data locality on a given host
   private val executorsByHost = new HashMap[String, HashSet[String]]
 
+  protected val hostsByRack = new HashMap[String, HashSet[String]]
+
   private val executorIdToHost = new HashMap[String, String]
 
   // Listener object to pass upcalls into
@@ -223,6 +225,9 @@ private[spark] class TaskSchedulerImpl(
         executorAdded(o.executorId, o.host)
         newExecAvail = true
       }
+      for (rack <- getRackForHost(o.host)) {
+        hostsByRack.getOrElseUpdate(rack, new HashSet[String]()) += o.host
+      }
     }
 
     // Randomly shuffle offers to avoid always placing tasks on the same set of workers.
@@ -418,6 +423,12 @@ private[spark] class TaskSchedulerImpl(
     execs -= executorId
     if (execs.isEmpty) {
       executorsByHost -= host
+      for (rack <- getRackForHost(host); hosts <- hostsByRack.get(rack)) {
+        hosts -= host
+        if (hosts.isEmpty) {
+          hostsByRack -= rack
+        }
+      }
     }
     executorIdToHost -= executorId
     rootPool.executorLost(executorId, host)
@@ -435,6 +446,10 @@ private[spark] class TaskSchedulerImpl(
     executorsByHost.contains(host)
   }
 
+  def hasHostAliveOnRack(rack: String): Boolean = synchronized {
+    hostsByRack.contains(rack)
+  }
+
   def isExecutorAlive(execId: String): Boolean = synchronized {
     activeExecutorIds.contains(execId)
   }
diff --git a/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala b/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala
index 059cc9085a..3bdc71d93b 100644
--- a/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala
+++ b/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala
@@ -191,7 +191,9 @@ private[spark] class TaskSetManager(
       addTo(pendingTasksForHost.getOrElseUpdate(loc.host, new ArrayBuffer))
       for (rack <- sched.getRackForHost(loc.host)) {
         addTo(pendingTasksForRack.getOrElseUpdate(rack, new ArrayBuffer))
-        hadAliveLocations = true
+        if(sched.hasHostAliveOnRack(rack)){
+          hadAliveLocations = true
+        }
       }
     }
 
@@ -748,7 +750,8 @@ private[spark] class TaskSetManager(
         pendingTasksForHost.keySet.exists(sched.hasExecutorsAliveOnHost(_))) {
       levels += NODE_LOCAL
     }
-    if (!pendingTasksForRack.isEmpty && getLocalityWait(RACK_LOCAL) != 0) {
+    if (!pendingTasksForRack.isEmpty && getLocalityWait(RACK_LOCAL) != 0 &&
+        pendingTasksForRack.keySet.exists(sched.hasHostAliveOnRack(_))) {
       levels += RACK_LOCAL
     }
     levels += ANY
@@ -761,7 +764,8 @@ private[spark] class TaskSetManager(
     def newLocAvail(index: Int): Boolean = {
       for (loc <- tasks(index).preferredLocations) {
         if (sched.hasExecutorsAliveOnHost(loc.host) ||
-            sched.getRackForHost(loc.host).isDefined) {
+           (sched.getRackForHost(loc.host).isDefined &&
+           sched.hasHostAliveOnRack(sched.getRackForHost(loc.host).get))) {
           return true
         }
       }
-- 
cgit v1.2.3