org.apache.hadoop.hbase.replication.regionserver

Class ReplicationSinkManager

java.lang.Object

org.apache.hadoop.hbase.replication.regionserver.ReplicationSinkManager

@InterfaceAudience.Private
public class ReplicationSinkManager
extends Object

Maintains a collection of peers to replicate to, and randomly selects a single peer to replicate to per set of data to replicate. Also handles keeping track of peer availability.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ReplicationSinkManager.SinkPeer Wraps a replication region server sink to provide the ability to identify it.

Field Summary

Fields

Modifier and Type	Field and Description
private Map<ServerName,Integer>	badReportCounts
private int	badSinkThreshold
private Connection	conn
(package private) static int	DEFAULT_BAD_SINK_THRESHOLD Default maximum number of times a replication sink can be reported as bad before it will no longer be provided as a sink for replication without the pool of replication sinks being refreshed.
(package private) static float	DEFAULT_REPLICATION_SOURCE_RATIO Default ratio of the total number of peer cluster region servers to consider replicating to.
private HBaseReplicationEndpoint	endpoint
private long	lastUpdateToPeers
private static org.slf4j.Logger	LOG
private String	peerClusterId
private Random	random
private float	ratio
private List<ServerName>	sinks

Constructor Summary

Constructors

Constructor and Description
ReplicationSinkManager(ClusterConnection conn, String peerClusterId, HBaseReplicationEndpoint endpoint, org.apache.hadoop.conf.Configuration conf) Instantiate for a single replication peer cluster.

Method Summary

Modifier and Type	Method and Description
void	chooseSinks() Refresh the list of sinks.
int	getNumSinks()
ReplicationSinkManager.SinkPeer	getReplicationSink() Get a randomly-chosen replication sink to replicate to.
protected List<ServerName>	getSinksForTesting()
void	reportBadSink(ReplicationSinkManager.SinkPeer sinkPeer) Report a SinkPeer as being bad (i.e.
void	reportSinkSuccess(ReplicationSinkManager.SinkPeer sinkPeer) Report that a SinkPeer successfully replicated a chunk of data.

Methods inherited from class java.lang.Object

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

Field Detail

LOG

private static final org.slf4j.Logger LOG

DEFAULT_BAD_SINK_THRESHOLD

static final int DEFAULT_BAD_SINK_THRESHOLD

Default maximum number of times a replication sink can be reported as bad before it will no longer be provided as a sink for replication without the pool of replication sinks being refreshed.

See Also:

Constant Field Values

DEFAULT_REPLICATION_SOURCE_RATIO

static final float DEFAULT_REPLICATION_SOURCE_RATIO

Default ratio of the total number of peer cluster region servers to consider replicating to.

See Also:

Constant Field Values

conn

private final Connection conn

peerClusterId

private final String peerClusterId

endpoint

private final HBaseReplicationEndpoint endpoint

badReportCounts

private final Map<ServerName,Integer> badReportCounts

ratio

private final float ratio

badSinkThreshold

private final int badSinkThreshold

random

private final Random random

lastUpdateToPeers

private long lastUpdateToPeers

sinks

private List<ServerName> sinks

Constructor Detail

ReplicationSinkManager

public ReplicationSinkManager(ClusterConnection conn,
                              String peerClusterId,
                              HBaseReplicationEndpoint endpoint,
                              org.apache.hadoop.conf.Configuration conf)

Instantiate for a single replication peer cluster.

Parameters:

conn - connection to the peer cluster

peerClusterId - identifier of the peer cluster

endpoint - replication endpoint for inter cluster replication

conf - HBase configuration, used for determining replication source ratio and bad peer threshold

Method Detail

getReplicationSink

public ReplicationSinkManager.SinkPeer getReplicationSink()
                                                   throws IOException

Get a randomly-chosen replication sink to replicate to.

Returns:

a replication sink to replicate to

Throws:

IOException

reportBadSink

public void reportBadSink(ReplicationSinkManager.SinkPeer sinkPeer)

Report a SinkPeer as being bad (i.e. an attempt to replicate to it failed). If a single SinkPeer is reported as bad more than replication.bad.sink.threshold times, it will be removed from the pool of potential replication targets. n * The SinkPeer that had a failed replication attempt on it

reportSinkSuccess

public void reportSinkSuccess(ReplicationSinkManager.SinkPeer sinkPeer)

Report that a SinkPeer successfully replicated a chunk of data. n * The SinkPeer that had a failed replication attempt on it

chooseSinks

public void chooseSinks()

Refresh the list of sinks.

getNumSinks

public int getNumSinks()

getSinksForTesting

protected List<ServerName> getSinksForTesting()