hadoop 2.6 AbstractYarnScheduler源代码解析

AbstractYarnScheduler　扩展了AbstractService 实现了 ResourceScheduler 。

SchedulerNode是AbstractYarnScheduler的一个参数化构建类，代表一个计算结点的各种属性，如下：

availableResource：可用的资源。

usedResource：已经使用的资源;

totalResourceCapability：总共资源容量

reservedContainer：保留的资源

numContainers：正在运行的容器数。

private final Map<ContainerId, RMContainer> launchedContainers = new HashMap<ContainerId, RMContainer>();//已经启动了的容器映射。

private final RMNode rmNode;//Node Manager的相关信息

private final String nodeName;//结点名称

private volatile Set<String> labels = null; //结点标签

RMNode 是一个接口，有以下内容：

public interface RMNode {

/** negative value means no timeout */
public static final int OVER_COMMIT_TIMEOUT_MILLIS_DEFAULT = -1;//如果此值为负，代表永不超时。

/**
* the node id of of this node.
* @return the node id of this node.
*/
public NodeId getNodeID();//结点ID

/**
* the hostname of this node
* @return hostname of this node
*/
public String getHostName();//主机名

/**
* the command port for this node
* @return command port for this node
*/
public int getCommandPort();//命令端口

/**
* the http port for this node
* @return http port for this node
*/
public int getHttpPort();//Http端口

/**
* the ContainerManager address for this node.
* @return the ContainerManager address for this node.
*/
public String getNodeAddress();//结点的ContainerManager地址

/**
* the http-Address for this node.
* @return the http-url address for this node
*/
public String getHttpAddress();//结点的http地址

/**
* the latest health report received from this node.
* @return the latest health report received from this node.
*/
public String getHealthReport();//健康状态报告

/**
* the time of the latest health report received from this node.
* @return the time of the latest health report received from this node.
*/
public long getLastHealthReportTime();//最后一次收到的健康报告时间

/**
* the node manager version of the node received as part of the
* registration with the resource manager
*/
public String getNodeManagerVersion();//node manager的版本

/**
* the total available resource.
* @return the total available resource.
*/
public Resource getTotalCapability();

/**
* The rack name for this node manager.
* @return the rack name.
*/
public String getRackName();

/**
* the {@link Node} information for this node.
* @return {@link Node} information for this node.
*/
public Node getNode();

public NodeState getState();

public List<ContainerId> getContainersToCleanUp();

public List<ApplicationId> getAppsToCleanup();

/**
* Update a {@link NodeHeartbeatResponse} with the list of containers and
* applications to clean up for this node.
* @param response the {@link NodeHeartbeatResponse} to update
*/
public void updateNodeHeartbeatResponseForCleanup(NodeHeartbeatResponse response);

public NodeHeartbeatResponse getLastNodeHeartBeatResponse();

/**
* Reset lastNodeHeartbeatResponse's ID to 0.
*/
void resetLastNodeHeartBeatResponse();

/**
* Get and clear the list of containerUpdates accumulated across NM
* heartbeats.
*
* @return containerUpdates accumulated across NM heartbeats.
*/
public List<UpdatedContainerInfo> pullContainerUpdates();

/**
* Get set of labels in this node
*
* @return labels in this node
*/
public Set<String> getNodeLabels();
}

NodeState是结点状态的枚举

public enum NodeState {
/** New node */
NEW,

/** Running node */
RUNNING,

/** Node is unhealthy */
UNHEALTHY,

/** Node is out of service */
DECOMMISSIONED,

/** Node has not sent a heartbeat for some configured time threshold*/
LOST,

/** Node has rebooted */
REBOOTED;

public boolean isUnusable() {
return (this == UNHEALTHY || this == DECOMMISSIONED || this == LOST);
}
}

NodeHeartbeatResponse

public interface NodeHeartbeatResponse {
int getResponseId();
NodeAction getNodeAction(); NodeAction是一个枚举，有三种状态  NORMAL, RESYNC, SHUTDOWN

List<ContainerId> getContainersToCleanup();
List<ContainerId> getContainersToBeRemovedFromNM();

List<ApplicationId> getApplicationsToCleanup();

void setResponseId(int responseId);
void setNodeAction(NodeAction action);

MasterKey getContainerTokenMasterKey();
void setContainerTokenMasterKey(MasterKey secretKey);

MasterKey getNMTokenMasterKey();
void setNMTokenMasterKey(MasterKey secretKey);

void addAllContainersToCleanup(List<ContainerId> containers);

// This tells NM to remove finished containers from its context. Currently, NM
// will remove finished containers from its context only after AM has actually
// received the finished containers in a previous allocate response
void addContainersToBeRemovedFromNM(List<ContainerId> containers);

void addAllApplicationsToCleanup(List<ApplicationId> applications);

long getNextHeartBeatInterval();
void setNextHeartBeatInterval(long nextHeartBeatInterval);

String getDiagnosticsMessage();

void setDiagnosticsMessage(String diagnosticsMessage);

// Credentials (i.e. hdfs tokens) needed by NodeManagers for application
// localizations and logAggreations.
Map<ApplicationId, ByteBuffer> getSystemCredentialsForApps();

void setSystemCredentialsForApps(
Map<ApplicationId, ByteBuffer> systemCredentials);
}

SchedulerApplication 有以下三个主要字段：

private Queue queue;
private final String user;
private T currentAttempt;