HDFS-8131. Implement a space balanced block placement policy. Contributed by Liu Shaohui.

This commit is contained in:
Kihwal Lee 2015-05-19 08:04:38 -05:00
parent eb4c9ddeb5
commit de30d66b26
5 changed files with 279 additions and 3 deletions

View File

@ -345,6 +345,9 @@ Release 2.8.0 - UNRELEASED
HDFS-7891. A block placement policy with best rack failure tolerance.
(Walter Su via szetszwo)
HDFS-8131. Implement a space balanced block placement policy (Liu Shaohui
via kihwal)
IMPROVEMENTS
HDFS-3918. EditLogTailer shouldn't log WARN when other node

View File

@ -519,7 +519,11 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
public static final String DFS_WEB_AUTHENTICATION_KERBEROS_KEYTAB_KEY = "dfs.web.authentication.kerberos.keytab";
public static final String DFS_NAMENODE_MAX_OP_SIZE_KEY = "dfs.namenode.max.op.size";
public static final int DFS_NAMENODE_MAX_OP_SIZE_DEFAULT = 50 * 1024 * 1024;
public static final String DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY =
"dfs.namenode.available-space-block-placement-policy.balanced-space-preference-fraction";
public static final float DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_DEFAULT =
0.6f;
public static final String DFS_BLOCK_LOCAL_PATH_ACCESS_USER_KEY = "dfs.block.local-path-access.user";
public static final String DFS_DOMAIN_SOCKET_PATH_KEY = "dfs.domain.socket.path";
public static final String DFS_DOMAIN_SOCKET_PATH_DEFAULT = "";

View File

@ -0,0 +1,95 @@
/**
* 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.
*/
package org.apache.hadoop.hdfs.server.blockmanagement;
import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY;
import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_DEFAULT;
import java.util.Random;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hdfs.DFSConfigKeys;
import org.apache.hadoop.net.NetworkTopology;
/**
* Space balanced block placement policy.
*/
public class AvailableSpaceBlockPlacementPolicy extends
BlockPlacementPolicyDefault {
private static final Log LOG = LogFactory
.getLog(AvailableSpaceBlockPlacementPolicy.class);
private static final Random RAND = new Random();
private int balancedPreference =
(int) (100 * DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_DEFAULT);
@Override
public void initialize(Configuration conf, FSClusterStats stats,
NetworkTopology clusterMap, Host2NodesMap host2datanodeMap) {
super.initialize(conf, stats, clusterMap, host2datanodeMap);
float balancedPreferencePercent =
conf.getFloat(
DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY,
DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_DEFAULT);
LOG.info("Available space block placement policy initialized: "
+ DFSConfigKeys.DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY
+ " = " + balancedPreferencePercent);
if (balancedPreferencePercent > 1.0) {
LOG.warn("The value of "
+ DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY
+ " is greater than 1.0 but should be in the range 0.0 - 1.0");
}
if (balancedPreferencePercent < 0.5) {
LOG.warn("The value of "
+ DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY
+ " is less than 0.5 so datanodes with more used percent will"
+ " receive more block allocations.");
}
balancedPreference = (int) (100 * balancedPreferencePercent);
}
@Override
protected DatanodeDescriptor chooseDataNode(String scope) {
DatanodeDescriptor a = (DatanodeDescriptor) clusterMap.chooseRandom(scope);
DatanodeDescriptor b = (DatanodeDescriptor) clusterMap.chooseRandom(scope);
int ret = compareDataNode(a, b);
if (ret == 0) {
return a;
} else if (ret < 0) {
return (RAND.nextInt(100) < balancedPreference) ? a : b;
} else {
return (RAND.nextInt(100) < balancedPreference) ? b : a;
}
}
/**
* Compare the two data nodes.
*/
protected int compareDataNode(final DatanodeDescriptor a,
final DatanodeDescriptor b) {
if (a.equals(b)
|| Math.abs(a.getDfsUsedPercent() - b.getDfsUsedPercent()) < 5) {
return 0;
}
return a.getDfsUsedPercent() < b.getDfsUsedPercent() ? -1 : 1;
}
}

View File

@ -646,8 +646,7 @@ protected DatanodeStorageInfo chooseRandom(int numOfReplicas,
boolean badTarget = false;
DatanodeStorageInfo firstChosen = null;
while(numOfReplicas > 0 && numOfAvailableNodes > 0) {
DatanodeDescriptor chosenNode =
(DatanodeDescriptor)clusterMap.chooseRandom(scope);
DatanodeDescriptor chosenNode = chooseDataNode(scope);
if (excludedNodes.add(chosenNode)) { //was not in the excluded list
if (LOG.isDebugEnabled()) {
builder.append("\nNode ").append(NodeBase.getPath(chosenNode)).append(" [");
@ -708,6 +707,14 @@ protected DatanodeStorageInfo chooseRandom(int numOfReplicas,
return firstChosen;
}
/**
* Choose a datanode from the given <i>scope</i>.
* @return the chosen node, if there is any.
*/
protected DatanodeDescriptor chooseDataNode(final String scope) {
return (DatanodeDescriptor) clusterMap.chooseRandom(scope);
}
/**
* If the given storage is a good target, add it to the result list and
* update the set of excluded nodes.

View File

@ -0,0 +1,167 @@
/**
* 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.
*/
package org.apache.hadoop.hdfs.server.blockmanagement;
import java.io.File;
import java.util.ArrayList;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.hdfs.DFSConfigKeys;
import org.apache.hadoop.hdfs.DFSTestUtil;
import org.apache.hadoop.hdfs.HdfsConfiguration;
import org.apache.hadoop.hdfs.TestBlockStoragePolicy;
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants;
import org.apache.hadoop.hdfs.server.namenode.NameNode;
import org.apache.hadoop.net.NetworkTopology;
import org.apache.hadoop.test.PathUtils;
import org.junit.AfterClass;
import org.junit.Assert;
import org.junit.BeforeClass;
import org.junit.Test;
public class TestAvailableSpaceBlockPlacementPolicy {
private final static int numRacks = 4;
private final static int nodesPerRack = 5;
private final static int blockSize = 1024;
private final static int chooseTimes = 10000;
private final static String file = "/tobers/test";
private final static int replica = 3;
private static DatanodeStorageInfo[] storages;
private static DatanodeDescriptor[] dataNodes;
private static Configuration conf;
private static NameNode namenode;
private static BlockPlacementPolicy placementPolicy;
private static NetworkTopology cluster;
@BeforeClass
public static void setupCluster() throws Exception {
conf = new HdfsConfiguration();
conf.setFloat(
DFSConfigKeys.DFS_NAMENODE_AVAILABLE_SPACE_BLOCK_PLACEMENT_POLICY_BALANCED_SPACE_PREFERENCE_FRACTION_KEY,
0.6f);
String[] racks = new String[numRacks];
for (int i = 0; i < numRacks; i++) {
racks[i] = "/rack" + i;
}
String[] owerRackOfNodes = new String[numRacks * nodesPerRack];
for (int i = 0; i < nodesPerRack; i++) {
for (int j = 0; j < numRacks; j++) {
owerRackOfNodes[i * numRacks + j] = racks[j];
}
}
storages = DFSTestUtil.createDatanodeStorageInfos(owerRackOfNodes);
dataNodes = DFSTestUtil.toDatanodeDescriptor(storages);
FileSystem.setDefaultUri(conf, "hdfs://localhost:0");
conf.set(DFSConfigKeys.DFS_NAMENODE_HTTP_ADDRESS_KEY, "0.0.0.0:0");
File baseDir = PathUtils.getTestDir(AvailableSpaceBlockPlacementPolicy.class);
conf.set(DFSConfigKeys.DFS_NAMENODE_NAME_DIR_KEY, new File(baseDir, "name").getPath());
conf.set(DFSConfigKeys.DFS_BLOCK_REPLICATOR_CLASSNAME_KEY,
AvailableSpaceBlockPlacementPolicy.class.getName());
DFSTestUtil.formatNameNode(conf);
namenode = new NameNode(conf);
final BlockManager bm = namenode.getNamesystem().getBlockManager();
placementPolicy = bm.getBlockPlacementPolicy();
cluster = bm.getDatanodeManager().getNetworkTopology();
for (int i = 0; i < nodesPerRack * numRacks; i++) {
cluster.add(dataNodes[i]);
}
setupDataNodeCapacity();
}
private static void updateHeartbeatWithUsage(DatanodeDescriptor dn,
long capacity, long dfsUsed, long remaining, long blockPoolUsed,
long dnCacheCapacity, long dnCacheUsed, int xceiverCount,
int volFailures) {
dn.getStorageInfos()[0].setUtilizationForTesting(
capacity, dfsUsed, remaining, blockPoolUsed);
dn.updateHeartbeat(
BlockManagerTestUtil.getStorageReportsForDatanode(dn),
dnCacheCapacity, dnCacheUsed, xceiverCount, volFailures, null);
}
private static void setupDataNodeCapacity() {
for (int i = 0; i < nodesPerRack * numRacks; i++) {
if ((i % 2) == 0) {
// remaining 100%
updateHeartbeatWithUsage(dataNodes[i], 2 * HdfsServerConstants.MIN_BLOCKS_FOR_WRITE * blockSize,
0L, 2 * HdfsServerConstants.MIN_BLOCKS_FOR_WRITE * blockSize, 0L, 0L, 0L, 0, 0);
} else {
// remaining 50%
updateHeartbeatWithUsage(dataNodes[i], 2 * HdfsServerConstants.MIN_BLOCKS_FOR_WRITE * blockSize,
HdfsServerConstants.MIN_BLOCKS_FOR_WRITE * blockSize, HdfsServerConstants.MIN_BLOCKS_FOR_WRITE
* blockSize, 0L, 0L, 0L, 0, 0);
}
}
}
/*
* To verify that the BlockPlacementPolicy can be replaced by AvailableSpaceBlockPlacementPolicy via
* changing the configuration.
*/
@Test
public void testPolicyReplacement() {
Assert.assertTrue((placementPolicy instanceof AvailableSpaceBlockPlacementPolicy));
}
/*
* Call choose target many times and verify that nodes with more remaining percent will be chosen
* with high possibility.
*/
@Test
public void testChooseTarget() {
int total = 0;
int moreRemainingNode = 0;
for (int i = 0; i < chooseTimes; i++) {
DatanodeStorageInfo[] targets =
namenode
.getNamesystem()
.getBlockManager()
.getBlockPlacementPolicy()
.chooseTarget(file, replica, null, new ArrayList<DatanodeStorageInfo>(), false, null,
blockSize, TestBlockStoragePolicy.DEFAULT_STORAGE_POLICY);
Assert.assertTrue(targets.length == replica);
for (int j = 0; j < replica; j++) {
total++;
if (targets[j].getDatanodeDescriptor().getRemainingPercent() > 60) {
moreRemainingNode++;
}
}
}
Assert.assertTrue(total == replica * chooseTimes);
double possibility = 1.0 * moreRemainingNode / total;
Assert.assertTrue(possibility > 0.52);
Assert.assertTrue(possibility < 0.55);
}
@AfterClass
public static void teardownCluster() {
if (namenode != null) {
namenode.stop();
}
}
}