org.apache.giraph.zk

Class ZooKeeperManager

java.lang.Object

org.apache.giraph.zk.ZooKeeperManager

public class ZooKeeperManager
extends Object

Manages the election of ZooKeeper servers, starting/stopping the services, etc.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ZooKeeperManager.State State of the application

Constructor Summary

Constructors

Constructor and Description
ZooKeeperManager(org.apache.hadoop.mapreduce.Mapper.Context context, ImmutableClassesGiraphConfiguration configuration) Constructor with context.

Method Summary

Modifier and Type	Method and Description
void	cleanup() Do necessary cleanup in zookeeper wrapper.
void	cleanupOnExit() Mark files zookeeper creates in hdfs to be deleted on exit.
boolean	computationDone() Check if all the computation is done.
void	createCandidateStamp() Create a HDFS stamp for this task.
static String	getBasePath(org.apache.hadoop.conf.Configuration conf) Return the base ZooKeeper ZNode from which all other ZNodes Giraph creates should be sited, for instance in a multi-tenant ZooKeeper, the znode reserved for Giraph
String	getZooKeeperServerPortString() Users can get the server port string to connect to ZooKeeper
void	offlineZooKeeperServers(ZooKeeperManager.State state) Notify the ZooKeeper servers that this partition is done with all ZooKeeper communication.
void	onlineZooKeeperServer() If this task has been selected, online a ZooKeeper server.
boolean	runsZooKeeper() Is this task running a ZooKeeper server? Only could be true if called after onlineZooKeeperServers().
void	setup() Create the candidate stamps and decide on the servers to start if you are partition 0.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ZooKeeperManager

public ZooKeeperManager(org.apache.hadoop.mapreduce.Mapper.Context context,
                        ImmutableClassesGiraphConfiguration configuration)
                 throws IOException

Constructor with context.

Parameters:

context - Context to be stored internally

configuration - Configuration

Throws:

IOException

Method Detail

getBasePath

public static String getBasePath(org.apache.hadoop.conf.Configuration conf)

Return the base ZooKeeper ZNode from which all other ZNodes Giraph creates should be sited, for instance in a multi-tenant ZooKeeper, the znode reserved for Giraph

Parameters:

conf - Necessary to access user-provided values

Returns:

String of path without trailing slash

setup

public void setup()
           throws IOException,
                  InterruptedException

Create the candidate stamps and decide on the servers to start if you are partition 0.

Throws:

IOException

InterruptedException

createCandidateStamp

public void createCandidateStamp()

Create a HDFS stamp for this task. If another task already created it, then this one will fail, which is fine.

computationDone

public boolean computationDone()

Check if all the computation is done.

Returns:

true if all computation is done.

getZooKeeperServerPortString

public String getZooKeeperServerPortString()

Users can get the server port string to connect to ZooKeeper

Returns:

server port string - comma separated

onlineZooKeeperServer

public void onlineZooKeeperServer()
                           throws IOException

If this task has been selected, online a ZooKeeper server. Otherwise, wait until this task knows that the ZooKeeper servers have been onlined.

Throws:

IOException

offlineZooKeeperServers

public void offlineZooKeeperServers(ZooKeeperManager.State state)

Notify the ZooKeeper servers that this partition is done with all ZooKeeper communication. If this task is running a ZooKeeper server, kill it when all partitions are done and wait for completion. Clean up the ZooKeeper local directory as well.

Parameters:

state - State of the application

runsZooKeeper

public boolean runsZooKeeper()

Is this task running a ZooKeeper server? Only could be true if called after onlineZooKeeperServers().

Returns:

true if running a ZooKeeper server, false otherwise

cleanupOnExit

public void cleanupOnExit()

Mark files zookeeper creates in hdfs to be deleted on exit. To be called on master, since it's the last one who finishes.

cleanup

public void cleanup()

Do necessary cleanup in zookeeper wrapper.