Author: tgraves
Date: Wed Aug 28 21:10:04 2013
New Revision: 1518385
URL: http://svn.apache.org/r1518385
Log:
YARN-1101. Active nodes can be decremented below 0 (Robert Parker via tgraves)
Modified:
hadoop/common/branches/branch-0.23/hadoop-yarn-project/CHANGES.txt
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMNodeTransitions.java
Modified: hadoop/common/branches/branch-0.23/hadoop-yarn-project/CHANGES.txt
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-0.23/hadoop-yarn-project/CHANGES.txt?rev=1518385&r1=1518384&r2=1518385&view=diff
==============================================================================
--- hadoop/common/branches/branch-0.23/hadoop-yarn-project/CHANGES.txt
(original)
+++ hadoop/common/branches/branch-0.23/hadoop-yarn-project/CHANGES.txt Wed Aug
28 21:10:04 2013
@@ -29,6 +29,9 @@ Release 0.23.10 - UNRELEASED
YARN-337. RM handles killed application tracking URL poorly (jlowe)
+ YARN-1101. Active nodes can be decremented below 0 (Robert Parker
+ via tgraves)
+
Release 0.23.9 - 2013-07-08
INCOMPATIBLE CHANGES
Modified:
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java?rev=1518385&r1=1518384&r2=1518385&view=diff
==============================================================================
---
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java
(original)
+++
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java
Wed Aug 28 21:10:04 2013
@@ -363,9 +363,18 @@ public class RMNodeImpl implements RMNod
}
}
- private void updateMetricsForDeactivatedNode(RMNodeState finalState) {
+ private void updateMetricsForDeactivatedNode(RMNodeState initialState,
+ RMNodeState finalState) {
ClusterMetrics metrics = ClusterMetrics.getMetrics();
- metrics.decrNumActiveNodes();
+
+ switch (initialState) {
+ case RUNNING:
+ metrics.decrNumActiveNodes();
+ break;
+ case UNHEALTHY:
+ metrics.decrNumUnhealthyNMs();
+ break;
+ }
switch (finalState) {
case DECOMMISSIONED:
@@ -480,7 +489,7 @@ public class RMNodeImpl implements RMNod
rmNode.context.getInactiveRMNodes().put(rmNode.nodeId.getHost(), rmNode);
//Update the metrics
- rmNode.updateMetricsForDeactivatedNode(finalState);
+ rmNode.updateMetricsForDeactivatedNode(rmNode.getState(), finalState);
}
}
@@ -504,7 +513,8 @@ public class RMNodeImpl implements RMNod
rmNode.context.getDispatcher().getEventHandler().handle(
new NodeRemovedSchedulerEvent(rmNode));
// Update metrics
- rmNode.updateMetricsForDeactivatedNode(RMNodeState.UNHEALTHY);
+ rmNode.updateMetricsForDeactivatedNode(rmNode.getState(),
+ RMNodeState.UNHEALTHY);
return RMNodeState.UNHEALTHY;
}
Modified:
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMNodeTransitions.java
URL:
http://svn.apache.org/viewvc/hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMNodeTransitions.java?rev=1518385&r1=1518384&r2=1518385&view=diff
==============================================================================
---
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMNodeTransitions.java
(original)
+++
hadoop/common/branches/branch-0.23/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMNodeTransitions.java
Wed Aug 28 21:10:04 2013
@@ -163,7 +163,21 @@ public class TestRMNodeTransitions {
@Test
public void testUnhealthyExpire() {
RMNodeImpl node = getUnhealthyNode();
+ ClusterMetrics cm = ClusterMetrics.getMetrics();
+ int initialActive = cm.getNumActiveNMs();
+ int initialLost = cm.getNumLostNMs();
+ int initialUnhealthy = cm.getUnhealthyNMs();
+ int initialDecommissioned = cm.getNumDecommisionedNMs();
+ int initialRebooted = cm.getNumRebootedNMs();
node.handle(new RMNodeEvent(node.getNodeID(), RMNodeEventType.EXPIRE));
+ Assert.assertEquals("Active Nodes", initialActive, cm.getNumActiveNMs());
+ Assert.assertEquals("Lost Nodes", initialLost + 1, cm.getNumLostNMs());
+ Assert.assertEquals("Unhealthy Nodes",
+ initialUnhealthy - 1, cm.getUnhealthyNMs());
+ Assert.assertEquals("Decommissioned Nodes",
+ initialDecommissioned, cm.getNumDecommisionedNMs());
+ Assert.assertEquals("Rebooted Nodes",
+ initialRebooted, cm.getNumRebootedNMs());
Assert.assertEquals(RMNodeState.LOST, node.getState());
}
@@ -178,8 +192,22 @@ public class TestRMNodeTransitions {
@Test
public void testUnhealthyDecommission() {
RMNodeImpl node = getUnhealthyNode();
+ ClusterMetrics cm = ClusterMetrics.getMetrics();
+ int initialActive = cm.getNumActiveNMs();
+ int initialLost = cm.getNumLostNMs();
+ int initialUnhealthy = cm.getUnhealthyNMs();
+ int initialDecommissioned = cm.getNumDecommisionedNMs();
+ int initialRebooted = cm.getNumRebootedNMs();
node.handle(new RMNodeEvent(node.getNodeID(),
RMNodeEventType.DECOMMISSION));
+ Assert.assertEquals("Active Nodes", initialActive, cm.getNumActiveNMs());
+ Assert.assertEquals("Lost Nodes", initialLost, cm.getNumLostNMs());
+ Assert.assertEquals("Unhealthy Nodes",
+ initialUnhealthy - 1, cm.getUnhealthyNMs());
+ Assert.assertEquals("Decommissioned Nodes",
+ initialDecommissioned + 1, cm.getNumDecommisionedNMs());
+ Assert.assertEquals("Rebooted Nodes",
+ initialRebooted, cm.getNumRebootedNMs());
Assert.assertEquals(RMNodeState.DECOMMISSIONED, node.getState());
}
@@ -194,8 +222,22 @@ public class TestRMNodeTransitions {
@Test
public void testUnhealthyRebooting() {
RMNodeImpl node = getUnhealthyNode();
+ ClusterMetrics cm = ClusterMetrics.getMetrics();
+ int initialActive = cm.getNumActiveNMs();
+ int initialLost = cm.getNumLostNMs();
+ int initialUnhealthy = cm.getUnhealthyNMs();
+ int initialDecommissioned = cm.getNumDecommisionedNMs();
+ int initialRebooted = cm.getNumRebootedNMs();
node.handle(new RMNodeEvent(node.getNodeID(),
RMNodeEventType.REBOOTING));
+ Assert.assertEquals("Active Nodes", initialActive, cm.getNumActiveNMs());
+ Assert.assertEquals("Lost Nodes", initialLost, cm.getNumLostNMs());
+ Assert.assertEquals("Unhealthy Nodes",
+ initialUnhealthy - 1, cm.getUnhealthyNMs());
+ Assert.assertEquals("Decommissioned Nodes",
+ initialDecommissioned, cm.getNumDecommisionedNMs());
+ Assert.assertEquals("Rebooted Nodes",
+ initialRebooted + 1, cm.getNumRebootedNMs());
Assert.assertEquals(RMNodeState.REBOOTED, node.getState());
}
