HBASE-17296 Provide per peer throttling for replication

Signed-off-by: Phil Yang <yangzhe1991@apache.org>
This commit is contained in:
Guanghao Zhang 2016-12-14 16:34:55 +08:00 committed by Phil Yang
parent e029c554bb
commit 4b3df0f926
15 changed files with 263 additions and 22 deletions

View File

@ -70,6 +70,12 @@ public interface ReplicationPeer {
*/
public Map<TableName, List<String>> getTableCFs();
/**
* Get the per node bandwidth upper limit for this peer
* @return the bandwidth up limit
*/
public long getPeerBandwidth();
/**
* Setup a callback for chanages to the replication peer config
* @param listener Listener for config changes, usually a replication endpoint

View File

@ -41,6 +41,7 @@ public class ReplicationPeerConfig {
private final Map<byte[], byte[]> peerData;
private final Map<String, String> configuration;
private Map<TableName, ? extends Collection<String>> tableCFsMap = null;
private long bandwidth = 0;
public ReplicationPeerConfig() {
this.peerData = new TreeMap<byte[], byte[]>(Bytes.BYTES_COMPARATOR);
@ -89,13 +90,23 @@ public class ReplicationPeerConfig {
this.tableCFsMap = tableCFsMap;
}
public long getBandwidth() {
return this.bandwidth;
}
public ReplicationPeerConfig setBandwidth(long bandwidth) {
this.bandwidth = bandwidth;
return this;
}
@Override
public String toString() {
StringBuilder builder = new StringBuilder("clusterKey=").append(clusterKey).append(",");
builder.append("replicationEndpointImpl=").append(replicationEndpointImpl);
if (tableCFsMap != null) {
builder.append(tableCFsMap.toString());
builder.append(tableCFsMap.toString()).append(",");
}
builder.append("bandwidth=").append(bandwidth);
return builder.toString();
}
}

View File

@ -163,6 +163,12 @@ public class ReplicationPeerZKImpl extends ReplicationStateZKBase implements Rep
return this.tableCFs;
}
@Override
public long getPeerBandwidth() {
return this.peerConfig.getBandwidth();
}
@Override
public void trackPeerConfigChanges(ReplicationPeerConfigListener listener) {
if (this.peerConfigTracker != null){

View File

@ -375,6 +375,7 @@ public class ReplicationPeersZKImpl extends ReplicationStateZKBase implements Re
// or data that weren't explicitly changed
existingConfig.getConfiguration().putAll(newConfig.getConfiguration());
existingConfig.getPeerData().putAll(newConfig.getPeerData());
existingConfig.setBandwidth(newConfig.getBandwidth());
try {
ZKUtil.setData(this.zookeeper, getPeerNode(id),
ReplicationSerDeHelper.toByteArray(existingConfig));

View File

@ -264,6 +264,10 @@ public final class ReplicationSerDeHelper {
if (tableCFsMap != null) {
peerConfig.setTableCFsMap(tableCFsMap);
}
if (peer.hasBandwidth()) {
peerConfig.setBandwidth(peer.getBandwidth());
}
return peerConfig;
}
@ -308,6 +312,7 @@ public final class ReplicationSerDeHelper {
}
}
builder.setBandwidth(peerConfig.getBandwidth());
return builder.build();
}
}

View File

@ -5867,6 +5867,16 @@ public final class ZooKeeperProtos {
*/
org.apache.hadoop.hbase.protobuf.generated.ZooKeeperProtos.TableCFOrBuilder getTableCfsOrBuilder(
int index);
// optional int64 bandwidth = 6;
/**
* <code>optional int64 bandwidth = 6;</code>
*/
boolean hasBandwidth();
/**
* <code>optional int64 bandwidth = 6;</code>
*/
long getBandwidth();
}
/**
* Protobuf type {@code hbase.pb.ReplicationPeer}
@ -5958,6 +5968,11 @@ public final class ZooKeeperProtos {
tableCfs_.add(input.readMessage(org.apache.hadoop.hbase.protobuf.generated.ZooKeeperProtos.TableCF.PARSER, extensionRegistry));
break;
}
case 48: {
bitField0_ |= 0x00000004;
bandwidth_ = input.readInt64();
break;
}
}
}
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
@ -6216,12 +6231,29 @@ public final class ZooKeeperProtos {
return tableCfs_.get(index);
}
// optional int64 bandwidth = 6;
public static final int BANDWIDTH_FIELD_NUMBER = 6;
private long bandwidth_;
/**
* <code>optional int64 bandwidth = 6;</code>
*/
public boolean hasBandwidth() {
return ((bitField0_ & 0x00000004) == 0x00000004);
}
/**
* <code>optional int64 bandwidth = 6;</code>
*/
public long getBandwidth() {
return bandwidth_;
}
private void initFields() {
clusterkey_ = "";
replicationEndpointImpl_ = "";
data_ = java.util.Collections.emptyList();
configuration_ = java.util.Collections.emptyList();
tableCfs_ = java.util.Collections.emptyList();
bandwidth_ = 0L;
}
private byte memoizedIsInitialized = -1;
public final boolean isInitialized() {
@ -6272,6 +6304,9 @@ public final class ZooKeeperProtos {
for (int i = 0; i < tableCfs_.size(); i++) {
output.writeMessage(5, tableCfs_.get(i));
}
if (((bitField0_ & 0x00000004) == 0x00000004)) {
output.writeInt64(6, bandwidth_);
}
getUnknownFields().writeTo(output);
}
@ -6301,6 +6336,10 @@ public final class ZooKeeperProtos {
size += com.google.protobuf.CodedOutputStream
.computeMessageSize(5, tableCfs_.get(i));
}
if (((bitField0_ & 0x00000004) == 0x00000004)) {
size += com.google.protobuf.CodedOutputStream
.computeInt64Size(6, bandwidth_);
}
size += getUnknownFields().getSerializedSize();
memoizedSerializedSize = size;
return size;
@ -6340,6 +6379,11 @@ public final class ZooKeeperProtos {
.equals(other.getConfigurationList());
result = result && getTableCfsList()
.equals(other.getTableCfsList());
result = result && (hasBandwidth() == other.hasBandwidth());
if (hasBandwidth()) {
result = result && (getBandwidth()
== other.getBandwidth());
}
result = result &&
getUnknownFields().equals(other.getUnknownFields());
return result;
@ -6373,6 +6417,10 @@ public final class ZooKeeperProtos {
hash = (37 * hash) + TABLE_CFS_FIELD_NUMBER;
hash = (53 * hash) + getTableCfsList().hashCode();
}
if (hasBandwidth()) {
hash = (37 * hash) + BANDWIDTH_FIELD_NUMBER;
hash = (53 * hash) + hashLong(getBandwidth());
}
hash = (29 * hash) + getUnknownFields().hashCode();
memoizedHashCode = hash;
return hash;
@ -6512,6 +6560,8 @@ public final class ZooKeeperProtos {
} else {
tableCfsBuilder_.clear();
}
bandwidth_ = 0L;
bitField0_ = (bitField0_ & ~0x00000020);
return this;
}
@ -6575,6 +6625,10 @@ public final class ZooKeeperProtos {
} else {
result.tableCfs_ = tableCfsBuilder_.build();
}
if (((from_bitField0_ & 0x00000020) == 0x00000020)) {
to_bitField0_ |= 0x00000004;
}
result.bandwidth_ = bandwidth_;
result.bitField0_ = to_bitField0_;
onBuilt();
return result;
@ -6679,6 +6733,9 @@ public final class ZooKeeperProtos {
}
}
}
if (other.hasBandwidth()) {
setBandwidth(other.getBandwidth());
}
this.mergeUnknownFields(other.getUnknownFields());
return this;
}
@ -7626,6 +7683,39 @@ public final class ZooKeeperProtos {
return tableCfsBuilder_;
}
// optional int64 bandwidth = 6;
private long bandwidth_ ;
/**
* <code>optional int64 bandwidth = 6;</code>
*/
public boolean hasBandwidth() {
return ((bitField0_ & 0x00000020) == 0x00000020);
}
/**
* <code>optional int64 bandwidth = 6;</code>
*/
public long getBandwidth() {
return bandwidth_;
}
/**
* <code>optional int64 bandwidth = 6;</code>
*/
public Builder setBandwidth(long value) {
bitField0_ |= 0x00000020;
bandwidth_ = value;
onChanged();
return this;
}
/**
* <code>optional int64 bandwidth = 6;</code>
*/
public Builder clearBandwidth() {
bitField0_ = (bitField0_ & ~0x00000020);
bandwidth_ = 0L;
onChanged();
return this;
}
// @@protoc_insertion_point(builder_scope:hbase.pb.ReplicationPeer)
}
@ -10915,24 +11005,24 @@ public final class ZooKeeperProtos {
"ENABLED\"?\n\005State\022\013\n\007ENABLED\020\000\022\014\n\010DISABLE" +
"D\020\001\022\r\n\tDISABLING\020\002\022\014\n\010ENABLING\020\003\"D\n\007Tabl" +
"eCF\022\'\n\ntable_name\030\001 \001(\0132\023.hbase.pb.Table" +
"Name\022\020\n\010families\030\002 \003(\014\"\305\001\n\017ReplicationPe" +
"Name\022\020\n\010families\030\002 \003(\014\"\330\001\n\017ReplicationPe" +
"er\022\022\n\nclusterkey\030\001 \002(\t\022\037\n\027replicationEnd" +
"pointImpl\030\002 \001(\t\022&\n\004data\030\003 \003(\0132\030.hbase.pb" +
".BytesBytesPair\022/\n\rconfiguration\030\004 \003(\0132\030" +
".hbase.pb.NameStringPair\022$\n\ttable_cfs\030\005 " +
"\003(\0132\021.hbase.pb.TableCF\"g\n\020ReplicationSta" +
"te\022/\n\005state\030\001 \002(\0162 .hbase.pb.Replication",
"State.State\"\"\n\005State\022\013\n\007ENABLED\020\000\022\014\n\010DIS" +
"ABLED\020\001\"+\n\027ReplicationHLogPosition\022\020\n\010po" +
"sition\030\001 \002(\003\"%\n\017ReplicationLock\022\022\n\nlock_" +
"owner\030\001 \002(\t\"\252\001\n\tTableLock\022\'\n\ntable_name\030" +
"\001 \001(\0132\023.hbase.pb.TableName\022(\n\nlock_owner" +
"\030\002 \001(\0132\024.hbase.pb.ServerName\022\021\n\tthread_i" +
"d\030\003 \001(\003\022\021\n\tis_shared\030\004 \001(\010\022\017\n\007purpose\030\005 " +
"\001(\t\022\023\n\013create_time\030\006 \001(\003\"\036\n\013SwitchState\022" +
"\017\n\007enabled\030\001 \001(\010BE\n*org.apache.hadoop.hb" +
"ase.protobuf.generatedB\017ZooKeeperProtosH",
"\001\210\001\001\240\001\001"
"\003(\0132\021.hbase.pb.TableCF\022\021\n\tbandwidth\030\006 \001(" +
"\003\"g\n\020ReplicationState\022/\n\005state\030\001 \002(\0162 .h",
"base.pb.ReplicationState.State\"\"\n\005State\022" +
"\013\n\007ENABLED\020\000\022\014\n\010DISABLED\020\001\"+\n\027Replicatio" +
"nHLogPosition\022\020\n\010position\030\001 \002(\003\"%\n\017Repli" +
"cationLock\022\022\n\nlock_owner\030\001 \002(\t\"\252\001\n\tTable" +
"Lock\022\'\n\ntable_name\030\001 \001(\0132\023.hbase.pb.Tabl" +
"eName\022(\n\nlock_owner\030\002 \001(\0132\024.hbase.pb.Ser" +
"verName\022\021\n\tthread_id\030\003 \001(\003\022\021\n\tis_shared\030" +
"\004 \001(\010\022\017\n\007purpose\030\005 \001(\t\022\023\n\013create_time\030\006 " +
"\001(\003\"\036\n\013SwitchState\022\017\n\007enabled\030\001 \001(\010BE\n*o" +
"rg.apache.hadoop.hbase.protobuf.generate",
"dB\017ZooKeeperProtosH\001\210\001\001\240\001\001"
};
com.google.protobuf.Descriptors.FileDescriptor.InternalDescriptorAssigner assigner =
new com.google.protobuf.Descriptors.FileDescriptor.InternalDescriptorAssigner() {
@ -10986,7 +11076,7 @@ public final class ZooKeeperProtos {
internal_static_hbase_pb_ReplicationPeer_fieldAccessorTable = new
com.google.protobuf.GeneratedMessage.FieldAccessorTable(
internal_static_hbase_pb_ReplicationPeer_descriptor,
new java.lang.String[] { "Clusterkey", "ReplicationEndpointImpl", "Data", "Configuration", "TableCfs", });
new java.lang.String[] { "Clusterkey", "ReplicationEndpointImpl", "Data", "Configuration", "TableCfs", "Bandwidth", });
internal_static_hbase_pb_ReplicationState_descriptor =
getDescriptor().getMessageTypes().get(8);
internal_static_hbase_pb_ReplicationState_fieldAccessorTable = new

View File

@ -135,6 +135,7 @@ message ReplicationPeer {
repeated BytesBytesPair data = 3;
repeated NameStringPair configuration = 4;
repeated TableCF table_cfs = 5;
optional int64 bandwidth = 6;
}
/**

View File

@ -62,6 +62,7 @@ import org.apache.hadoop.hbase.regionserver.wal.WALEdit;
import org.apache.hadoop.hbase.replication.ChainWALEntryFilter;
import org.apache.hadoop.hbase.replication.ReplicationEndpoint;
import org.apache.hadoop.hbase.replication.ReplicationException;
import org.apache.hadoop.hbase.replication.ReplicationPeer;
import org.apache.hadoop.hbase.replication.ReplicationPeers;
import org.apache.hadoop.hbase.replication.ReplicationQueueInfo;
import org.apache.hadoop.hbase.replication.ReplicationQueues;
@ -144,6 +145,8 @@ public class ReplicationSource extends Thread
private WALEntryFilter walEntryFilter;
// throttler
private ReplicationThrottler throttler;
private long defaultBandwidth;
private long currentBandwidth;
private ConcurrentHashMap<String, ReplicationSourceWorkerThread> workerThreads =
new ConcurrentHashMap<String, ReplicationSourceWorkerThread>();
@ -179,8 +182,6 @@ public class ReplicationSource extends Thread
this.maxRetriesMultiplier =
this.conf.getInt("replication.source.maxretriesmultiplier", 300); // 5 minutes @ 1 sec per
this.queueSizePerGroup = this.conf.getInt("hbase.regionserver.maxlogs", 32);
long bandwidth = this.conf.getLong("replication.source.per.peer.node.bandwidth", 0);
this.throttler = new ReplicationThrottler((double)bandwidth/10.0);
this.replicationQueues = replicationQueues;
this.replicationPeers = replicationPeers;
this.manager = manager;
@ -196,6 +197,15 @@ public class ReplicationSource extends Thread
this.actualPeerId = replicationQueueInfo.getPeerId();
this.logQueueWarnThreshold = this.conf.getInt("replication.source.log.queue.warn", 2);
this.replicationEndpoint = replicationEndpoint;
defaultBandwidth = this.conf.getLong("replication.source.per.peer.node.bandwidth", 0);
currentBandwidth = getCurrentBandwidth();
this.throttler = new ReplicationThrottler((double) currentBandwidth / 10.0);
LOG.info("peerClusterZnode=" + peerClusterZnode + ", ReplicationSource : " + peerId
+ " inited, replicationQueueSizeCapacity=" + replicationQueueSizeCapacity
+ ", replicationQueueNbCapacity=" + replicationQueueNbCapacity + ", curerntBandwidth="
+ this.currentBandwidth);
}
private void decorateConf() {
@ -494,6 +504,13 @@ public class ReplicationSource extends Thread
return this.metrics;
}
private long getCurrentBandwidth() {
ReplicationPeer replicationPeer = this.replicationPeers.getPeer(peerId);
long peerBandwidth = replicationPeer != null ? replicationPeer.getPeerBandwidth() : 0;
// user can set peer bandwidth to 0 to use default bandwidth
return peerBandwidth != 0 ? peerBandwidth : defaultBandwidth;
}
public class ReplicationSourceWorkerThread extends Thread {
private ReplicationSource source;
private String walGroupId;
@ -1087,6 +1104,16 @@ public class ReplicationSource extends Thread
return distinctRowKeys + totalHFileEntries;
}
private void checkBandwidthChangeAndResetThrottler() {
long peerBandwidth = getCurrentBandwidth();
if (peerBandwidth != currentBandwidth) {
currentBandwidth = peerBandwidth;
throttler.setBandwidth((double) currentBandwidth / 10.0);
LOG.info("ReplicationSource : " + peerId
+ " bandwidth throttling changed, currentBandWidth=" + currentBandwidth);
}
}
/**
* Do the shipping logic
* @param currentWALisBeingWrittenTo was the current WAL being (seemingly)
@ -1101,6 +1128,7 @@ public class ReplicationSource extends Thread
}
while (isWorkerActive()) {
try {
checkBandwidthChangeAndResetThrottler();
if (throttler.isEnabled()) {
long sleepTicks = throttler.getNextSleepInterval(currentSize);
if (sleepTicks > 0) {

View File

@ -27,8 +27,8 @@ import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
*/
@InterfaceAudience.Private
public class ReplicationThrottler {
private final boolean enabled;
private final double bandwidth;
private boolean enabled;
private double bandwidth;
private long cyclePushSize;
private long cycleStartTick;
@ -118,4 +118,9 @@ public class ReplicationThrottler {
this.cycleStartTick = EnvironmentEdgeManager.currentTime();
}
}
public void setBandwidth(double bandwidth) {
this.bandwidth = bandwidth;
this.enabled = this.bandwidth > 0;
}
}

View File

@ -363,4 +363,21 @@ public class TestReplicationAdmin {
assertNull(admin.getPeerConfig(ID_ONE).getTableCFsMap());
admin.removePeer(ID_ONE);
}
@Test
public void testPeerBandwidth() throws ReplicationException {
ReplicationPeerConfig rpc = new ReplicationPeerConfig();
rpc.setClusterKey(KEY_ONE);
admin.addPeer(ID_ONE, rpc);
admin.peerAdded(ID_ONE);
rpc = admin.getPeerConfig(ID_ONE);
assertEquals(0, rpc.getBandwidth());
rpc.setBandwidth(2097152);
admin.updatePeerConfig(ID_ONE, rpc);
assertEquals(2097152, admin.getPeerConfig(ID_ONE).getBandwidth());
admin.removePeer(ID_ONE);
}
}

View File

@ -182,6 +182,15 @@ module Hbase
@replication_admin.removePeerTableCFs(id, map)
end
# Set new bandwidth config for the specified peer
def set_peer_bandwidth(id, bandwidth)
rpc = get_peer_config(id)
unless rpc.nil?
rpc.setBandwidth(bandwidth)
@replication_admin.updatePeerConfig(id, rpc)
end
end
#----------------------------------------------------------------------------------------------
# Enables a table's replication switch
def enable_tablerep(table_name)

View File

@ -359,6 +359,7 @@ Shell.load_command_group(
disable_peer
show_peer_tableCFs
set_peer_tableCFs
set_peer_bandwidth
list_replicated_tables
append_peer_tableCFs
remove_peer_tableCFs

View File

@ -34,13 +34,14 @@ EOF
peers = replication_admin.list_peers
formatter.header(["PEER_ID", "CLUSTER_KEY", "ENDPOINT_CLASSNAME",
"STATE", "TABLE_CFS"])
"STATE", "TABLE_CFS", "BANDWIDTH"])
peers.entrySet().each do |e|
state = replication_admin.get_peer_state(e.key)
tableCFs = replication_admin.show_peer_tableCFs(e.key)
formatter.row([ e.key, e.value.getClusterKey,
e.value.getReplicationEndpointImpl, state, tableCFs ])
e.value.getReplicationEndpointImpl, state, tableCFs,
e.value.getBandwidth ])
end
formatter.footer(now)

View File

@ -0,0 +1,42 @@
#
# Copyright The Apache Software Foundation
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
module Shell
module Commands
class SetPeerBandwidth< Command
def help
return <<-EOF
Set the replication source per node bandwidth for the specified peer.
Examples:
# set bandwidth=2MB per regionserver for a peer
hbase> set_peer_bandwidth '1', 2097152
# unset bandwidth for a peer to use the default bandwidth configured in server-side
hbase> set_peer_bandwidth '1'
EOF
end
def command(id, bandwidth = 0)
replication_admin.set_peer_bandwidth(id, bandwidth)
end
end
end
end

View File

@ -198,6 +198,24 @@ module Hbase
replication_admin.remove_peer(@peer_id)
end
define_test "set_peer_bandwidth: works with peer bandwidth upper limit" do
cluster_key = "localhost:2181:/hbase-test"
args = { CLUSTER_KEY => cluster_key }
replication_admin.add_peer(@peer_id, args)
# Normally the ReplicationSourceManager will call ReplicationPeer#peer_added
# but here we have to do it ourselves
replication_admin.peer_added(@peer_id)
peer_config = replication_admin.get_peer_config(@peer_id)
assert_equal(0, peer_config.get_bandwidth)
replication_admin.set_peer_bandwidth(@peer_id, 2097152)
peer_config = replication_admin.get_peer_config(@peer_id)
assert_equal(2097152, peer_config.get_bandwidth)
#cleanup
replication_admin.remove_peer(@peer_id)
end
define_test "get_peer_config: works with simple clusterKey peer" do
cluster_key = "localhost:2181:/hbase-test"
args = { CLUSTER_KEY => cluster_key }