Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

MAPREDUCE-3360. Added information about lost/rebooted/decommissioned nodes on the webapps. Contributed by Bhallamudi Venkata Siva Kamesh and Jason Lowe. 

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1236433 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Vinod Kumar Vavilapalli 2012-01-26 22:56:53 +00:00
parent 54f738575c
commit 3cdc100369
20 changed files with 390 additions and 134 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
hadoop-mapreduce-project/CHANGES.txt
View File

@ -229,6 +229,9 @@ Release 0.23.1 - Unreleased
MAPREDUCE-3718. Change default AM heartbeat interval to 1 second. (Hitesh
Shah via sseth)
MAPREDUCE-3360. Added information about lost/rebooted/decommissioned nodes
on the webapps. (Bhallamudi Venkata Siva Kamesh and Jason Lowe via vinodkv)
BUG FIXES
MAPREDUCE-3221. Reenabled the previously ignored test in TestSubmitJob

23
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ClusterMetrics.java
View File

@ -29,7 +29,6 @@ import org.apache.hadoop.metrics2.annotation.Metric;
import org.apache.hadoop.metrics2.annotation.Metrics;
import org.apache.hadoop.metrics2.lib.DefaultMetricsSystem;
import org.apache.hadoop.metrics2.lib.MetricsRegistry;
import org.apache.hadoop.metrics2.lib.MutableCounterInt;
import org.apache.hadoop.metrics2.lib.MutableGaugeInt;
import org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeEventType;
@ -39,9 +38,9 @@ public class ClusterMetrics {
private static AtomicBoolean isInitialized = new AtomicBoolean(false);
@Metric("# of NMs") MutableGaugeInt numNMs;
@Metric("# of decommissioned NMs") MutableCounterInt numDecommissionedNMs;
@Metric("# of lost NMs") MutableCounterInt numLostNMs;
@Metric("# of active NMs") MutableGaugeInt numNMs;
@Metric("# of decommissioned NMs") MutableGaugeInt numDecommissionedNMs;
@Metric("# of lost NMs") MutableGaugeInt numLostNMs;
@Metric("# of unhealthy NMs") MutableGaugeInt numUnhealthyNMs;
@Metric("# of Rebooted NMs") MutableGaugeInt numRebootedNMs;
@ -73,8 +72,8 @@ public class ClusterMetrics {
}
}
//Total Nodemanagers
public int getNumNMs() {
//Active Nodemanagers
public int getNumActiveNMs() {
return numNMs.value();
}
@ -87,6 +86,10 @@ public class ClusterMetrics {
numDecommissionedNMs.incr();
}
public void decrDecommisionedNMs() {
numDecommissionedNMs.decr();
}
//Lost NMs
public int getNumLostNMs() {
return numLostNMs.value();
@ -96,6 +99,10 @@ public class ClusterMetrics {
numLostNMs.incr();
}
public void decrNumLostNMs() {
numLostNMs.decr();
}
//Unhealthy NMs
public int getUnhealthyNMs() {
return numUnhealthyNMs.value();
@ -118,6 +125,10 @@ public class ClusterMetrics {
numRebootedNMs.incr();
}
public void decrNumRebootedNMs() {
numRebootedNMs.decr();
}
public void removeNode(RMNodeEventType nodeEventType) {
numNMs.decr();
switch(nodeEventType){

2
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMContext.java
View File

@ -44,6 +44,8 @@ public interface RMContext {
ConcurrentMap<ApplicationId, RMApp> getRMApps();
ConcurrentMap<String, RMNode> getInactiveRMNodes();
ConcurrentMap<NodeId, RMNode> getRMNodes();
AMLivelinessMonitor getAMLivelinessMonitor();

8
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMContextImpl.java
View File

@ -44,6 +44,9 @@ public class RMContextImpl implements RMContext {
private final ConcurrentMap<NodeId, RMNode> nodes
= new ConcurrentHashMap<NodeId, RMNode>();
private final ConcurrentMap<String, RMNode> inactiveNodes
= new ConcurrentHashMap<String, RMNode>();
private AMLivelinessMonitor amLivelinessMonitor;
private ContainerAllocationExpirer containerAllocationExpirer;
private final DelegationTokenRenewer tokenRenewer;
@ -84,6 +87,11 @@ public class RMContextImpl implements RMContext {
return this.nodes;
}
@Override
public ConcurrentMap<String, RMNode> getInactiveRMNodes() {
return this.inactiveNodes;
}
@Override
public ContainerAllocationExpirer getContainerAllocationExpirer() {
return this.containerAllocationExpirer;

4
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceTrackerService.java
View File

@ -220,10 +220,6 @@ public class ResourceTrackerService extends AbstractService implements
if (rmNode == null) {
/* node does not exist */
LOG.info("Node not found rebooting " + remoteNodeStatus.getNodeId());
// Updating the metrics directly as reboot event cannot be
// triggered on a null rmNode
ClusterMetrics.getMetrics().incrNumRebootedNMs();
return reboot;
}

26
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmnode/RMNodeImpl.java
View File

@ -119,7 +119,7 @@ public class RMNodeImpl implements RMNode, EventHandler<RMNodeEvent> {
RMNodeEventType.DECOMMISSION, new RemoveNodeTransition())
.addTransition(RMNodeState.RUNNING, RMNodeState.LOST,
RMNodeEventType.EXPIRE, new RemoveNodeTransition())
.addTransition(RMNodeState.RUNNING, RMNodeState.LOST,
.addTransition(RMNodeState.RUNNING, RMNodeState.REBOOTED,
RMNodeEventType.REBOOTING, new RemoveNodeTransition())
.addTransition(RMNodeState.RUNNING, RMNodeState.RUNNING,
RMNodeEventType.CLEANUP_APP, new CleanUpAppTransition())
@ -308,6 +308,21 @@ public class RMNodeImpl implements RMNode, EventHandler<RMNodeEvent> {
public static class AddNodeTransition implements
SingleArcTransition<RMNodeImpl, RMNodeEvent> {
private void updateMetrics(RMNodeState nodeState) {
ClusterMetrics metrics = ClusterMetrics.getMetrics();
switch (nodeState) {
case LOST:
metrics.decrNumLostNMs();
break;
case REBOOTED:
metrics.decrNumRebootedNMs();
break;
case DECOMMISSIONED:
metrics.decrDecommisionedNMs();
break;
}
}
@SuppressWarnings("unchecked")
@Override
public void transition(RMNodeImpl rmNode, RMNodeEvent event) {
@ -316,6 +331,13 @@ public class RMNodeImpl implements RMNode, EventHandler<RMNodeEvent> {
rmNode.context.getDispatcher().getEventHandler().handle(
new NodeAddedSchedulerEvent(rmNode));
String host = rmNode.nodeId.getHost();
if (rmNode.context.getInactiveRMNodes().containsKey(host)) {
RMNode node = rmNode.context.getInactiveRMNodes().get(host);
rmNode.context.getInactiveRMNodes().remove(host);
updateMetrics(node.getState());
}
ClusterMetrics.getMetrics().addNode();
}
}
@ -353,7 +375,7 @@ public class RMNodeImpl implements RMNode, EventHandler<RMNodeEvent> {
// Remove the node from the system.
rmNode.context.getRMNodes().remove(rmNode.nodeId);
LOG.info("Removed Node " + rmNode.nodeId);
rmNode.context.getInactiveRMNodes().put(rmNode.nodeId.getHost(), rmNode);
//Update the metrics
ClusterMetrics.getMetrics().removeNode(event.getType());
}

4
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/MetricsOverviewTable.java
View File

@ -68,7 +68,7 @@ public class MetricsOverviewTable extends HtmlBlock {
th().$class("ui-state-default")._("Memory Used")._().
th().$class("ui-state-default")._("Memory Total")._().
th().$class("ui-state-default")._("Memory Reserved")._().
th().$class("ui-state-default")._("Total Nodes")._().
th().$class("ui-state-default")._("Active Nodes")._().
th().$class("ui-state-default")._("Decommissioned Nodes")._().
th().$class("ui-state-default")._("Lost Nodes")._().
th().$class("ui-state-default")._("Unhealthy Nodes")._().
@ -82,7 +82,7 @@ public class MetricsOverviewTable extends HtmlBlock {
td(StringUtils.byteDesc(clusterMetrics.getAllocatedMB() * BYTES_IN_MB)).
td(StringUtils.byteDesc(clusterMetrics.getTotalMB() * BYTES_IN_MB)).
td(StringUtils.byteDesc(clusterMetrics.getReservedMB() * BYTES_IN_MB)).
td().a(url("nodes"),String.valueOf(clusterMetrics.getTotalNodes()))._().
td().a(url("nodes"),String.valueOf(clusterMetrics.getActiveNodes()))._().
td().a(url("nodes/decommissioned"),String.valueOf(clusterMetrics.getDecommissionedNodes()))._().
td().a(url("nodes/lost"),String.valueOf(clusterMetrics.getLostNodes()))._().
td().a(url("nodes/unhealthy"),String.valueOf(clusterMetrics.getUnhealthyNodes()))._().

30
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/NodesPage.java
View File

@ -24,6 +24,8 @@ import static org.apache.hadoop.yarn.webapp.view.JQueryUI.DATATABLES_ID;
import static org.apache.hadoop.yarn.webapp.view.JQueryUI.initID;
import static org.apache.hadoop.yarn.webapp.view.JQueryUI.tableInit;
import java.util.Collection;
import org.apache.hadoop.util.StringUtils;
import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
import org.apache.hadoop.yarn.server.resourcemanager.ResourceManager;
@ -36,6 +38,7 @@ import org.apache.hadoop.yarn.webapp.SubView;
import org.apache.hadoop.yarn.webapp.hamlet.Hamlet;
import org.apache.hadoop.yarn.webapp.hamlet.Hamlet.TABLE;
import org.apache.hadoop.yarn.webapp.hamlet.Hamlet.TBODY;
import org.apache.hadoop.yarn.webapp.hamlet.Hamlet.TR;
import org.apache.hadoop.yarn.webapp.view.HtmlBlock;
import com.google.inject.Inject;
@ -79,7 +82,19 @@ class NodesPage extends RmView {
if(type != null && !type.isEmpty()) {
stateFilter = RMNodeState.valueOf(type.toUpperCase());
}
for (RMNode ni : this.rmContext.getRMNodes().values()) {
Collection<RMNode> rmNodes = this.rmContext.getRMNodes().values();
boolean isInactive = false;
if (stateFilter != null) {
switch (stateFilter) {
case DECOMMISSIONED:
case LOST:
case REBOOTED:
rmNodes = this.rmContext.getInactiveRMNodes().values();
isInactive = true;
break;
}
}
for (RMNode ni : rmNodes) {
if(stateFilter != null) {
RMNodeState state = ni.getState();
if(!stateFilter.equals(state)) {
@ -89,12 +104,17 @@ class NodesPage extends RmView {
NodeInfo info = new NodeInfo(ni, sched);
int usedMemory = (int)info.getUsedMemory();
int availableMemory = (int)info.getAvailableMemory();
tbody.tr().
TR<TBODY<TABLE<Hamlet>>> row = tbody.tr().
td(info.getRack()).
td(info.getState()).
td(info.getNodeId()).
td().a("http://" + info.getNodeHTTPAddress(), info.getNodeHTTPAddress())._().
td(info.getHealthStatus()).
td(info.getNodeId());
if (isInactive) {
row.td()._("N/A")._();
} else {
String httpAddress = info.getNodeHTTPAddress();
row.td().a("http://" + httpAddress, httpAddress)._();
}
row.td(info.getHealthStatus()).
td(Times.format(info.getLastHealthUpdate())).
td(info.getHealthReport()).
td(String.valueOf(info.getNumContainers())).

33
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/RMWebServices.java
View File

@ -19,6 +19,7 @@
package org.apache.hadoop.yarn.server.resourcemanager.webapp;
import java.io.IOException;
import java.util.Collection;
import java.util.concurrent.ConcurrentMap;
import javax.servlet.http.HttpServletRequest;
@ -68,6 +69,7 @@ import com.google.inject.Singleton;
@Singleton
@Path("/ws/v1/cluster")
public class RMWebServices {
private static final String EMPTY = "";
private static final Log LOG = LogFactory.getLog(RMWebServices.class);
private final ResourceManager rm;
private static RecordFactory recordFactory = RecordFactoryProvider
@ -144,12 +146,23 @@ public class RMWebServices {
if (sched == null) {
throw new NotFoundException("Null ResourceScheduler instance");
}
Collection<RMNode> rmNodes = this.rm.getRMContext().getRMNodes().values();
boolean isInactive = false;
if (filterState != null && !filterState.isEmpty()) {
RMNodeState nodeState = RMNodeState.valueOf(filterState.toUpperCase());
switch (nodeState) {
case DECOMMISSIONED:
case LOST:
case REBOOTED:
rmNodes = this.rm.getRMContext().getInactiveRMNodes().values();
isInactive = true;
break;
}
}
NodesInfo allNodes = new NodesInfo();
for (RMNode ni : this.rm.getRMContext().getRMNodes().values()) {
for (RMNode ni : rmNodes) {
NodeInfo nodeInfo = new NodeInfo(ni, sched);
if (filterState != null) {
RMNodeState.valueOf(filterState);
if (!(nodeInfo.getState().equalsIgnoreCase(filterState))) {
continue;
}
@ -165,6 +178,9 @@ public class RMWebServices {
continue;
}
}
if (isInactive) {
nodeInfo.setNodeHTTPAddress(EMPTY);
}
allNodes.add(nodeInfo);
}
return allNodes;
@ -183,10 +199,19 @@ public class RMWebServices {
}
NodeId nid = ConverterUtils.toNodeId(nodeId);
RMNode ni = this.rm.getRMContext().getRMNodes().get(nid);
boolean isInactive = false;
if (ni == null) {
ni = this.rm.getRMContext().getInactiveRMNodes().get(nid.getHost());
if (ni == null) {
throw new NotFoundException("nodeId, " + nodeId + ", is not found");
}
return new NodeInfo(ni, sched);
isInactive = true;
}
NodeInfo nodeInfo = new NodeInfo(ni, sched);
if (isInactive) {
nodeInfo.setNodeHTTPAddress(EMPTY);
}
return nodeInfo;
}
@GET

10
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/dao/ClusterMetricsInfo.java
View File

@ -44,6 +44,7 @@ public class ClusterMetricsInfo {
protected int unhealthyNodes;
protected int decommissionedNodes;
protected int rebootedNodes;
protected int activeNodes;
public ClusterMetricsInfo() {
} // JAXB needs this
@ -59,12 +60,13 @@ public class ClusterMetricsInfo {
this.allocatedMB = metrics.getAllocatedGB() * MB_IN_GB;
this.containersAllocated = metrics.getAllocatedContainers();
this.totalMB = availableMB + reservedMB + allocatedMB;
this.totalNodes = clusterMetrics.getNumNMs();
this.activeNodes = clusterMetrics.getNumActiveNMs();
this.lostNodes = clusterMetrics.getNumLostNMs();
this.unhealthyNodes = clusterMetrics.getUnhealthyNMs();
this.decommissionedNodes = clusterMetrics.getNumDecommisionedNMs();
this.rebootedNodes = clusterMetrics.getNumRebootedNMs();
this.totalNodes = activeNodes + lostNodes + decommissionedNodes
+ rebootedNodes;
}
public int getAppsSubmitted() {
@ -95,6 +97,10 @@ public class ClusterMetricsInfo {
return this.totalNodes;
}
public int getActiveNodes() {
return this.activeNodes;
}
public int getLostNodes() {
return this.lostNodes;
}

4
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/dao/NodeInfo.java
View File

@ -95,6 +95,10 @@ public class NodeInfo {
return this.nodeHTTPAddress;
}
public void setNodeHTTPAddress(String nodeHTTPAddress) {
this.nodeHTTPAddress = nodeHTTPAddress;
}
public String getHealthStatus() {
return this.healthStatus;
}

10
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockNM.java
View File

@ -81,13 +81,20 @@ public class MockNM {
}
public HeartbeatResponse nodeHeartbeat(boolean b) throws Exception {
return nodeHeartbeat(new HashMap<ApplicationId, List<ContainerStatus>>(), b);
return nodeHeartbeat(new HashMap<ApplicationId, List<ContainerStatus>>(),
b, ++responseId);
}
public HeartbeatResponse nodeHeartbeat(Map<ApplicationId,
List<ContainerStatus>> conts, boolean isHealthy) throws Exception {
return nodeHeartbeat(conts, isHealthy, ++responseId);
}
public HeartbeatResponse nodeHeartbeat(Map<ApplicationId,
List<ContainerStatus>> conts, boolean isHealthy, int resId) throws Exception {
NodeHeartbeatRequest req = Records.newRecord(NodeHeartbeatRequest.class);
NodeStatus status = Records.newRecord(NodeStatus.class);
status.setResponseId(resId);
status.setNodeId(nodeId);
for (Map.Entry<ApplicationId, List<ContainerStatus>> entry : conts.entrySet()) {
status.setContainersStatuses(entry.getValue());
@ -97,7 +104,6 @@ public class MockNM {
healthStatus.setIsNodeHealthy(isHealthy);
healthStatus.setLastHealthReportTime(1);
status.setNodeHealthStatus(healthStatus);
status.setResponseId(++responseId);
req.setNodeStatus(status);
return resourceTracker.nodeHeartbeat(req).getHeartbeatResponse();
}

133
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockNodes.java
View File

@ -57,6 +57,17 @@ public class MockNodes {
return list;
}
public static List<RMNode> lostNodes(int racks, int nodesPerRack,
Resource perNode) {
List<RMNode> list = Lists.newArrayList();
for (int i = 0; i < racks; ++i) {
for (int j = 0; j < nodesPerRack; ++j) {
list.add(lostNodeInfo(i, perNode, RMNodeState.LOST));
}
}
return list;
}
public static NodeId newNodeID(String host, int port) {
NodeId nid = recordFactory.newRecordInstance(NodeId.class);
nid.setHost(host);
@ -82,41 +93,74 @@ public class MockNodes {
return rs;
}
public static RMNode newNodeInfo(int rack, final Resource perNode) {
final String rackName = "rack"+ rack;
final int nid = NODE_ID++;
final String hostName = "host"+ nid;
final int port = 123;
final NodeId nodeID = newNodeID(hostName, port);
final String httpAddress = "localhost:0";
final NodeHealthStatus nodeHealthStatus =
recordFactory.newRecordInstance(NodeHealthStatus.class);
final Resource used = newUsedResource(perNode);
final Resource avail = newAvailResource(perNode, used);
return new RMNode() {
private static class MockRMNodeImpl implements RMNode {
private NodeId nodeId;
private String hostName;
private String nodeAddr;
private String httpAddress;
private int cmdPort;
private Resource perNode;
private String rackName;
private NodeHealthStatus nodeHealthStatus;
private RMNodeState state;
public MockRMNodeImpl(NodeId nodeId, String nodeAddr, String httpAddress,
Resource perNode, String rackName, NodeHealthStatus nodeHealthStatus,
int cmdPort, String hostName, RMNodeState state) {
this.nodeId = nodeId;
this.nodeAddr = nodeAddr;
this.httpAddress = httpAddress;
this.perNode = perNode;
this.rackName = rackName;
this.nodeHealthStatus = nodeHealthStatus;
this.cmdPort = cmdPort;
this.hostName = hostName;
this.state = state;
}
@Override
public NodeId getNodeID() {
return nodeID;
return this.nodeId;
}
@Override
public String getHostName() {
return this.hostName;
}
@Override
public int getCommandPort() {
return this.cmdPort;
}
@Override
public int getHttpPort() {
return 0;
}
@Override
public String getNodeAddress() {
return hostName;
return this.nodeAddr;
}
@Override
public String getHttpAddress() {
return httpAddress;
return this.httpAddress;
}
@Override
public NodeHealthStatus getNodeHealthStatus() {
return this.nodeHealthStatus;
}
@Override
public Resource getTotalCapability() {
return perNode;
return this.perNode;
}
@Override
public String getRackName() {
return rackName;
return this.rackName;
}
@Override
@ -124,50 +168,45 @@ public class MockNodes {
throw new UnsupportedOperationException("Not supported yet.");
}
@Override
public NodeHealthStatus getNodeHealthStatus() {
return nodeHealthStatus;
}
@Override
public int getCommandPort() {
return nid;
}
@Override
public int getHttpPort() {
// TODO Auto-generated method stub
return 0;
}
@Override
public String getHostName() {
return hostName;
}
@Override
public RMNodeState getState() {
// TODO Auto-generated method stub
return this.state;
}
@Override
public List<ContainerId> getContainersToCleanUp() {
return null;
}
@Override
public List<ApplicationId> getAppsToCleanup() {
// TODO Auto-generated method stub
return null;
}
@Override
public List<ContainerId> getContainersToCleanUp() {
// TODO Auto-generated method stub
return null;
}
@Override
public HeartbeatResponse getLastHeartBeatResponse() {
// TODO Auto-generated method stub
return null;
}
};
private static RMNode buildRMNode(int rack, final Resource perNode, RMNodeState state, String httpAddr) {
final String rackName = "rack"+ rack;
final int nid = NODE_ID++;
final String hostName = "host"+ nid;
final int port = 123;
final NodeId nodeID = newNodeID(hostName, port);
final String httpAddress = httpAddr;
final NodeHealthStatus nodeHealthStatus =
recordFactory.newRecordInstance(NodeHealthStatus.class);
return new MockRMNodeImpl(nodeID, hostName, httpAddress, perNode, rackName,
nodeHealthStatus, nid, hostName, state);
}
public static RMNode lostNodeInfo(int rack, final Resource perNode, RMNodeState state) {
return buildRMNode(rack, perNode, state, "N/A");
}
public static RMNode newNodeInfo(int rack, final Resource perNode) {
return buildRMNode(rack, perNode, null, "localhost:0");
}
}

6
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockRM.java
View File

@ -131,6 +131,12 @@ public class MockRM extends ResourceManager {
node.handle(new RMNodeEvent(nm.getNodeId(), RMNodeEventType.STARTED));
}
public void sendNodeLost(MockNM nm) throws Exception {
RMNodeImpl node = (RMNodeImpl) getRMContext().getRMNodes().get(
nm.getNodeId());
node.handle(new RMNodeEvent(nm.getNodeId(), RMNodeEventType.EXPIRE));
}
public void NMwaitForState(NodeId nodeid, RMNodeState finalState)
throws Exception {
RMNode node = getRMContext().getRMNodes().get(nodeid);

5
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMNodeTransitions.java
View File

@ -31,6 +31,7 @@ import junit.framework.Assert;
import org.apache.hadoop.yarn.api.records.ContainerId;
import org.apache.hadoop.yarn.api.records.ContainerStatus;
import org.apache.hadoop.yarn.api.records.NodeHealthStatus;
import org.apache.hadoop.yarn.api.records.NodeId;
import org.apache.hadoop.yarn.event.EventHandler;
import org.apache.hadoop.yarn.server.api.records.HeartbeatResponse;
import org.apache.hadoop.yarn.server.resourcemanager.recovery.MemStore;
@ -100,8 +101,8 @@ public class TestRMNodeTransitions {
rmDispatcher.register(SchedulerEventType.class,
new TestSchedulerEventDispatcher());
node = new RMNodeImpl(null, rmContext, null, 0, 0, null, null);
NodeId nodeId = BuilderUtils.newNodeId("localhost", 0);
node = new RMNodeImpl(nodeId, rmContext, null, 0, 0, null, null);
}

4
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestResourceTrackerService.java
View File

@ -157,14 +157,14 @@ public class TestResourceTrackerService {
rm.start();
MockNM nm1 = rm.registerNode("host1:1234", 5120);
MockNM nm2 = new MockNM("host2:1234", 2048, rm.getResourceTrackerService());
MockNM nm2 = rm.registerNode("host2:1234", 2048);
int initialMetricCount = ClusterMetrics.getMetrics().getNumRebootedNMs();
HeartbeatResponse nodeHeartbeat = nm1.nodeHeartbeat(true);
Assert.assertTrue(NodeAction.NORMAL.equals(nodeHeartbeat.getNodeAction()));
nodeHeartbeat = nm2.nodeHeartbeat(
new HashMap<ApplicationId, List<ContainerStatus>>(), true);
new HashMap<ApplicationId, List<ContainerStatus>>(), true, -100);
Assert.assertTrue(NodeAction.REBOOT.equals(nodeHeartbeat.getNodeAction()));
checkRebootedNMCount(rm, ++initialMetricCount);
}

49
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/TestNodesPage.java
View File

@ -24,6 +24,7 @@ import org.apache.hadoop.yarn.server.resourcemanager.RMContext;
import org.apache.hadoop.yarn.server.resourcemanager.ResourceManager;
import org.apache.hadoop.yarn.server.resourcemanager.webapp.NodesPage.NodesBlock;
import org.apache.hadoop.yarn.webapp.test.WebAppTests;
import org.junit.Before;
import org.junit.Test;
import org.mockito.Mockito;
@ -37,8 +38,6 @@ import com.google.inject.Module;
*/
public class TestNodesPage {
@Test
public void testNodesBlockRender() throws Exception {
final int numberOfRacks = 2;
final int numberOfNodesPerRack = 2;
// Number of Actual Table Headers for NodesPage.NodesBlock might change in
@ -46,29 +45,57 @@ public class TestNodesPage {
final int numberOfThInMetricsTable = 10;
final int numberOfActualTableHeaders = 10;
Injector injector = WebAppTests.createMockInjector(RMContext.class,
TestRMWebApp.mockRMContext(3, numberOfRacks, numberOfNodesPerRack, 8*TestRMWebApp.GiB),
new Module() {
private Injector injector;
@Before
public void setUp() throws Exception {
injector = WebAppTests.createMockInjector(RMContext.class, TestRMWebApp
.mockRMContext(3, numberOfRacks, numberOfNodesPerRack,
8 * TestRMWebApp.GiB), new Module() {
@Override
public void configure(Binder binder) {
try {
binder.bind(ResourceManager.class).toInstance(TestRMWebApp.mockRm(3,
numberOfRacks, numberOfNodesPerRack, 8*TestRMWebApp.GiB));
binder.bind(ResourceManager.class).toInstance(
TestRMWebApp.mockRm(3, numberOfRacks, numberOfNodesPerRack,
8 * TestRMWebApp.GiB));
} catch (IOException e) {
throw new IllegalStateException(e);
}
}
});
}
@Test
public void testNodesBlockRender() throws Exception {
injector.getInstance(NodesBlock.class).render();
PrintWriter writer = injector.getInstance(PrintWriter.class);
WebAppTests.flushOutput(injector);
Mockito.verify(writer, Mockito.times(numberOfActualTableHeaders +
numberOfThInMetricsTable)).print(
"<th");
Mockito.verify(writer,
Mockito.times(numberOfActualTableHeaders + numberOfThInMetricsTable))
.print("<th");
Mockito.verify(
writer,
Mockito.times(numberOfRacks * numberOfNodesPerRack
* numberOfActualTableHeaders + numberOfThInMetricsTable)).print("<td");
* numberOfActualTableHeaders + numberOfThInMetricsTable)).print(
"<td");
}
@Test
public void testNodesBlockRenderForLostNodes() {
NodesBlock nodesBlock = injector.getInstance(NodesBlock.class);
nodesBlock.set("node.state", "lost");
nodesBlock.render();
PrintWriter writer = injector.getInstance(PrintWriter.class);
WebAppTests.flushOutput(injector);
Mockito.verify(writer,
Mockito.times(numberOfActualTableHeaders + numberOfThInMetricsTable))
.print("<th");
Mockito.verify(
writer,
Mockito.times(numberOfRacks * numberOfNodesPerRack
* numberOfActualTableHeaders + numberOfThInMetricsTable)).print(
"<td");
}
}

11
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/TestRMWebApp.java
View File

@ -120,12 +120,23 @@ public class TestRMWebApp {
for (RMNode node : nodes) {
nodesMap.put(node.getNodeID(), node);
}
final List<RMNode> lostNodes = MockNodes.lostNodes(racks, numNodes,
newResource(mbsPerNode));
final ConcurrentMap<String, RMNode> lostNodesMap = Maps.newConcurrentMap();
for (RMNode node : lostNodes) {
lostNodesMap.put(node.getHostName(), node);
}
return new RMContextImpl(new MemStore(), null, null, null, null) {
@Override
public ConcurrentMap<ApplicationId, RMApp> getRMApps() {
return applicationsMaps;
}
@Override
public ConcurrentMap<String, RMNode> getInactiveRMNodes() {
return lostNodesMap;
}
@Override
public ConcurrentMap<NodeId, RMNode> getRMNodes() {
return nodesMap;
}

18
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/TestRMWebServices.java
View File

@ -370,7 +370,8 @@ public class TestRMWebServices extends JerseyTest {
WebServicesTestUtils.getXmlInt(element, "lostNodes"),
WebServicesTestUtils.getXmlInt(element, "unhealthyNodes"),
WebServicesTestUtils.getXmlInt(element, "decommissionedNodes"),
WebServicesTestUtils.getXmlInt(element, "rebootedNodes"));
WebServicesTestUtils.getXmlInt(element, "rebootedNodes"),
WebServicesTestUtils.getXmlInt(element, "activeNodes"));
}
}
@ -378,7 +379,7 @@ public class TestRMWebServices extends JerseyTest {
Exception {
assertEquals("incorrect number of elements", 1, json.length());
JSONObject clusterinfo = json.getJSONObject("clusterMetrics");
assertEquals("incorrect number of elements", 11, clusterinfo.length());
assertEquals("incorrect number of elements", 12, clusterinfo.length());
verifyClusterMetrics(clusterinfo.getInt("appsSubmitted"),
clusterinfo.getInt("reservedMB"), clusterinfo.getInt("availableMB"),
clusterinfo.getInt("allocatedMB"),
@ -386,13 +387,13 @@ public class TestRMWebServices extends JerseyTest {
clusterinfo.getInt("totalMB"), clusterinfo.getInt("totalNodes"),
clusterinfo.getInt("lostNodes"), clusterinfo.getInt("unhealthyNodes"),
clusterinfo.getInt("decommissionedNodes"),
clusterinfo.getInt("rebootedNodes"));
clusterinfo.getInt("rebootedNodes"),clusterinfo.getInt("activeNodes"));
}
public void verifyClusterMetrics(int sub, int reservedMB, int availableMB,
int allocMB, int containersAlloc, int totalMB, int totalNodes,
int lostNodes, int unhealthyNodes, int decommissionedNodes,
int rebootedNodes) throws JSONException, Exception {
int rebootedNodes, int activeNodes) throws JSONException, Exception {
ResourceScheduler rs = rm.getResourceScheduler();
QueueMetrics metrics = rs.getRootQueueMetrics();
@ -412,8 +413,11 @@ public class TestRMWebServices extends JerseyTest {
* MB_IN_GB, allocMB);
assertEquals("containersAllocated doesn't match", 0, containersAlloc);
assertEquals("totalMB doesn't match", totalMBExpect, totalMB);
assertEquals("totalNodes doesn't match", clusterMetrics.getNumNMs(),
totalNodes);
assertEquals(
"totalNodes doesn't match",
clusterMetrics.getNumActiveNMs() + clusterMetrics.getNumLostNMs()
+ clusterMetrics.getNumDecommisionedNMs()
+ clusterMetrics.getNumRebootedNMs(), totalNodes);
assertEquals("lostNodes doesn't match", clusterMetrics.getNumLostNMs(),
lostNodes);
assertEquals("unhealthyNodes doesn't match",
@ -422,6 +426,8 @@ public class TestRMWebServices extends JerseyTest {
clusterMetrics.getNumDecommisionedNMs(), decommissionedNodes);
assertEquals("rebootedNodes doesn't match",
clusterMetrics.getNumRebootedNMs(), rebootedNodes);
assertEquals("activeNodes doesn't match", clusterMetrics.getNumActiveNMs(),
activeNodes);
}
@Test

63
hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/webapp/TestRMWebServicesNodes.java
View File

@ -203,6 +203,69 @@ public class TestRMWebServicesNodes extends JerseyTest {
}
}
@Test
public void testNodesQueryStateLost() throws JSONException, Exception {
WebResource r = resource();
MockNM nm1 = rm.registerNode("h1:1234", 5120);
MockNM nm2 = rm.registerNode("h2:1234", 5120);
rm.sendNodeStarted(nm1);
rm.sendNodeStarted(nm2);
rm.NMwaitForState(nm1.getNodeId(), RMNodeState.RUNNING);
rm.NMwaitForState(nm2.getNodeId(), RMNodeState.RUNNING);
rm.sendNodeLost(nm1);
rm.sendNodeLost(nm2);
ClientResponse response = r.path("ws").path("v1").path("cluster")
.path("nodes").queryParam("state", RMNodeState.LOST.toString())
.accept(MediaType.APPLICATION_JSON).get(ClientResponse.class);
assertEquals(MediaType.APPLICATION_JSON_TYPE, response.getType());
JSONObject json = response.getEntity(JSONObject.class);
JSONObject nodes = json.getJSONObject("nodes");
assertEquals("incorrect number of elements", 1, nodes.length());
JSONArray nodeArray = nodes.getJSONArray("node");
assertEquals("incorrect number of elements", 2, nodeArray.length());
for (int i = 0; i < nodeArray.length(); ++i) {
JSONObject info = nodeArray.getJSONObject(i);
String host = info.get("id").toString().split(":")[0];
RMNode rmNode = rm.getRMContext().getInactiveRMNodes().get(host);
WebServicesTestUtils.checkStringMatch("nodeHTTPAddress", "",
info.getString("nodeHTTPAddress"));
WebServicesTestUtils.checkStringMatch("state", rmNode.getState()
.toString(), info.getString("state"));
}
}
@Test
public void testSingleNodeQueryStateLost() throws JSONException, Exception {
WebResource r = resource();
MockNM nm1 = rm.registerNode("h1:1234", 5120);
MockNM nm2 = rm.registerNode("h2:1234", 5120);
rm.sendNodeStarted(nm1);
rm.sendNodeStarted(nm2);
rm.NMwaitForState(nm1.getNodeId(), RMNodeState.RUNNING);
rm.NMwaitForState(nm2.getNodeId(), RMNodeState.RUNNING);
rm.sendNodeLost(nm1);
rm.sendNodeLost(nm2);
ClientResponse response = r.path("ws").path("v1").path("cluster")
.path("nodes").path("h2:1234").accept(MediaType.APPLICATION_JSON)
.get(ClientResponse.class);
assertEquals(MediaType.APPLICATION_JSON_TYPE, response.getType());
JSONObject json = response.getEntity(JSONObject.class);
JSONObject info = json.getJSONObject("node");
String id = info.get("id").toString();
assertEquals("Incorrect Node Information.", "h2:1234", id);
RMNode rmNode = rm.getRMContext().getInactiveRMNodes().get("h2");
WebServicesTestUtils.checkStringMatch("nodeHTTPAddress", "",
info.getString("nodeHTTPAddress"));
WebServicesTestUtils.checkStringMatch("state",
rmNode.getState().toString(), info.getString("state"));
}
@Test
public void testNodesQueryHealthy() throws JSONException, Exception {
WebResource r = resource();