HDFS-6478. RemoteException can't be retried properly for non-HA scenario. Contributed by Ming Ma.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1611410 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Jing Zhao 2014-07-17 17:37:22 +00:00
parent 0a02b5a19b
commit f4151bbf4f
6 changed files with 71 additions and 71 deletions

View File

@ -321,6 +321,9 @@ Release 2.6.0 - UNRELEASED
HDFS-6689. NFS doesn't return correct lookup access for direcories (brandonli)
HDFS-6478. RemoteException can't be retried properly for non-HA scenario.
(Ming Ma via jing9)
Release 2.5.0 - UNRELEASED
INCOMPATIBLE CHANGES

View File

@ -333,19 +333,18 @@ private static NamenodeProtocol createNNProxyWithNamenodeProtocol(
address, conf, ugi, NamenodeProtocolPB.class, 0);
if (withRetries) { // create the proxy with retries
RetryPolicy timeoutPolicy = RetryPolicies.exponentialBackoffRetry(5, 200,
TimeUnit.MILLISECONDS);
Map<Class<? extends Exception>, RetryPolicy> exceptionToPolicyMap
= new HashMap<Class<? extends Exception>, RetryPolicy>();
RetryPolicy methodPolicy = RetryPolicies.retryByException(timeoutPolicy,
exceptionToPolicyMap);
Map<String, RetryPolicy> methodNameToPolicyMap
= new HashMap<String, RetryPolicy>();
methodNameToPolicyMap.put("getBlocks", methodPolicy);
methodNameToPolicyMap.put("getAccessKeys", methodPolicy);
proxy = (NamenodeProtocolPB) RetryProxy.create(NamenodeProtocolPB.class,
proxy, methodNameToPolicyMap);
TimeUnit.MILLISECONDS);
Map<String, RetryPolicy> methodNameToPolicyMap
= new HashMap<String, RetryPolicy>();
methodNameToPolicyMap.put("getBlocks", timeoutPolicy);
methodNameToPolicyMap.put("getAccessKeys", timeoutPolicy);
NamenodeProtocol translatorProxy =
new NamenodeProtocolTranslatorPB(proxy);
return (NamenodeProtocol) RetryProxy.create(
NamenodeProtocol.class, translatorProxy, methodNameToPolicyMap);
} else {
return new NamenodeProtocolTranslatorPB(proxy);
}
return new NamenodeProtocolTranslatorPB(proxy);
}
private static ClientProtocol createNNProxyWithClientProtocol(
@ -379,29 +378,27 @@ private static ClientProtocol createNNProxyWithClientProtocol(
= new HashMap<Class<? extends Exception>, RetryPolicy>();
remoteExceptionToPolicyMap.put(AlreadyBeingCreatedException.class,
createPolicy);
Map<Class<? extends Exception>, RetryPolicy> exceptionToPolicyMap
= new HashMap<Class<? extends Exception>, RetryPolicy>();
exceptionToPolicyMap.put(RemoteException.class, RetryPolicies
.retryByRemoteException(defaultPolicy,
remoteExceptionToPolicyMap));
RetryPolicy methodPolicy = RetryPolicies.retryByException(
defaultPolicy, exceptionToPolicyMap);
RetryPolicy methodPolicy = RetryPolicies.retryByRemoteException(
defaultPolicy, remoteExceptionToPolicyMap);
Map<String, RetryPolicy> methodNameToPolicyMap
= new HashMap<String, RetryPolicy>();
methodNameToPolicyMap.put("create", methodPolicy);
proxy = (ClientNamenodeProtocolPB) RetryProxy.create(
ClientNamenodeProtocolPB.class,
new DefaultFailoverProxyProvider<ClientNamenodeProtocolPB>(
ClientNamenodeProtocolPB.class, proxy),
ClientProtocol translatorProxy =
new ClientNamenodeProtocolTranslatorPB(proxy);
return (ClientProtocol) RetryProxy.create(
ClientProtocol.class,
new DefaultFailoverProxyProvider<ClientProtocol>(
ClientProtocol.class, translatorProxy),
methodNameToPolicyMap,
defaultPolicy);
} else {
return new ClientNamenodeProtocolTranslatorPB(proxy);
}
return new ClientNamenodeProtocolTranslatorPB(proxy);
}
private static Object createNameNodeProxy(InetSocketAddress address,
Configuration conf, UserGroupInformation ugi, Class<?> xface,
int rpcTimeout) throws IOException {

View File

@ -97,7 +97,7 @@ public DatanodeProtocolClientSideTranslatorPB(InetSocketAddress nameNodeAddr,
RPC.setProtocolEngine(conf, DatanodeProtocolPB.class,
ProtobufRpcEngine.class);
UserGroupInformation ugi = UserGroupInformation.getCurrentUser();
rpcProxy = createNamenodeWithRetry(createNamenode(nameNodeAddr, conf, ugi));
rpcProxy = createNamenode(nameNodeAddr, conf, ugi);
}
private static DatanodeProtocolPB createNamenode(
@ -109,33 +109,6 @@ private static DatanodeProtocolPB createNamenode(
org.apache.hadoop.ipc.Client.getPingInterval(conf), null).getProxy();
}
/** Create a {@link NameNode} proxy */
static DatanodeProtocolPB createNamenodeWithRetry(
DatanodeProtocolPB rpcNamenode) {
RetryPolicy createPolicy = RetryPolicies
.retryUpToMaximumCountWithFixedSleep(5,
HdfsConstants.LEASE_SOFTLIMIT_PERIOD, TimeUnit.MILLISECONDS);
Map<Class<? extends Exception>, RetryPolicy> remoteExceptionToPolicyMap =
new HashMap<Class<? extends Exception>, RetryPolicy>();
remoteExceptionToPolicyMap.put(AlreadyBeingCreatedException.class,
createPolicy);
Map<Class<? extends Exception>, RetryPolicy> exceptionToPolicyMap =
new HashMap<Class<? extends Exception>, RetryPolicy>();
exceptionToPolicyMap.put(RemoteException.class, RetryPolicies
.retryByRemoteException(RetryPolicies.TRY_ONCE_THEN_FAIL,
remoteExceptionToPolicyMap));
RetryPolicy methodPolicy = RetryPolicies.retryByException(
RetryPolicies.TRY_ONCE_THEN_FAIL, exceptionToPolicyMap);
Map<String, RetryPolicy> methodNameToPolicyMap = new HashMap<String, RetryPolicy>();
methodNameToPolicyMap.put("create", methodPolicy);
return (DatanodeProtocolPB) RetryProxy.create(DatanodeProtocolPB.class,
rpcNamenode, methodNameToPolicyMap);
}
@Override
public void close() throws IOException {
RPC.stopProxy(rpcProxy);

View File

@ -47,6 +47,7 @@
import org.apache.hadoop.hdfs.server.protocol.RemoteEditLogManifest;
import org.apache.hadoop.ipc.ProtobufHelper;
import org.apache.hadoop.ipc.ProtocolMetaInterface;
import org.apache.hadoop.ipc.ProtocolTranslator;
import org.apache.hadoop.ipc.RPC;
import org.apache.hadoop.ipc.RpcClientUtil;
@ -61,7 +62,7 @@
@InterfaceAudience.Private
@InterfaceStability.Stable
public class NamenodeProtocolTranslatorPB implements NamenodeProtocol,
ProtocolMetaInterface, Closeable {
ProtocolMetaInterface, Closeable, ProtocolTranslator {
/** RpcController is not used and hence is set to null */
private final static RpcController NULL_CONTROLLER = null;
@ -88,6 +89,11 @@ public void close() {
RPC.stopProxy(rpcProxy);
}
@Override
public Object getUnderlyingProxyObject() {
return rpcProxy;
}
@Override
public BlocksWithLocations getBlocks(DatanodeInfo datanode, long size)
throws IOException {

View File

@ -30,6 +30,8 @@
import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_REPLICATION_MIN_KEY;
import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_REPLICATION_DEFAULT;
import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_REPLICATION_KEY;
import static org.apache.hadoop.test.MetricsAsserts.assertCounter;
import static org.apache.hadoop.test.MetricsAsserts.getMetrics;
import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertTrue;
import static org.junit.Assert.fail;
@ -79,6 +81,7 @@
import org.apache.hadoop.hdfs.server.protocol.NamenodeProtocols;
import org.apache.hadoop.io.EnumSetWritable;
import org.apache.hadoop.io.IOUtils;
import org.apache.hadoop.metrics2.MetricsRecordBuilder;
import org.apache.hadoop.security.UserGroupInformation;
import org.apache.hadoop.test.GenericTestUtils;
import org.apache.hadoop.util.Time;
@ -97,6 +100,8 @@ public class TestFileCreation {
((Log4JLogger)LogFactory.getLog(FSNamesystem.class)).getLogger().setLevel(Level.ALL);
((Log4JLogger)DFSClient.LOG).getLogger().setLevel(Level.ALL);
}
private static final String RPC_DETAILED_METRICS =
"RpcDetailedActivityForPort";
static final long seed = 0xDEADBEEFL;
static final int blockSize = 8192;
@ -371,7 +376,7 @@ public void testOverwriteOpenForWrite() throws Exception {
conf.setBoolean(DFSConfigKeys.DFS_PERMISSIONS_ENABLED_KEY, false);
final MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).build();
FileSystem fs = cluster.getFileSystem();
UserGroupInformation otherUgi = UserGroupInformation.createUserForTesting(
"testuser", new String[]{"testgroup"});
FileSystem fs2 = otherUgi.doAs(new PrivilegedExceptionAction<FileSystem>() {
@ -380,12 +385,16 @@ public FileSystem run() throws Exception {
return FileSystem.get(cluster.getConfiguration(0));
}
});
String metricsName = RPC_DETAILED_METRICS + cluster.getNameNodePort();
try {
Path p = new Path("/testfile");
FSDataOutputStream stm1 = fs.create(p);
stm1.write(1);
assertCounter("CreateNumOps", 1L, getMetrics(metricsName));
// Create file again without overwrite
try {
fs2.create(p, false);
@ -394,7 +403,9 @@ public FileSystem run() throws Exception {
GenericTestUtils.assertExceptionContains("already being created by",
abce);
}
// NameNodeProxies' createNNProxyWithClientProtocol has 5 retries.
assertCounter("AlreadyBeingCreatedExceptionNumOps",
6L, getMetrics(metricsName));
FSDataOutputStream stm2 = fs2.create(p, true);
stm2.write(2);
stm2.close();

View File

@ -25,14 +25,16 @@
import org.apache.hadoop.hdfs.protocol.ClientProtocol;
import org.apache.hadoop.hdfs.protocolPB.ClientDatanodeProtocolTranslatorPB;
import org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB;
import org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolPB;
import org.apache.hadoop.hdfs.protocolPB.DatanodeProtocolClientSideTranslatorPB;
import org.apache.hadoop.hdfs.protocolPB.InterDatanodeProtocolTranslatorPB;
import org.apache.hadoop.hdfs.protocolPB.JournalProtocolTranslatorPB;
import org.apache.hadoop.hdfs.protocolPB.NamenodeProtocolTranslatorPB;
import org.apache.hadoop.hdfs.protocolPB.NamenodeProtocolPB;
import org.apache.hadoop.hdfs.server.datanode.DataNode;
import org.apache.hadoop.hdfs.server.protocol.JournalProtocol;
import org.apache.hadoop.hdfs.server.protocol.NamenodeProtocol;
import org.apache.hadoop.ipc.RPC;
import org.apache.hadoop.ipc.RpcClientUtil;
import org.apache.hadoop.net.NetUtils;
import org.apache.hadoop.security.RefreshUserMappingsProtocol;
import org.apache.hadoop.security.UserGroupInformation;
@ -76,16 +78,22 @@ public static void tearDown() throws Exception {
@Test
public void testNamenodeProtocol() throws IOException {
NamenodeProtocolTranslatorPB translator =
(NamenodeProtocolTranslatorPB) NameNodeProxies.createNonHAProxy(conf,
NamenodeProtocol np =
NameNodeProxies.createNonHAProxy(conf,
nnAddress, NamenodeProtocol.class, UserGroupInformation.getCurrentUser(),
true).getProxy();
boolean exists = translator.isMethodSupported("rollEditLog");
boolean exists = RpcClientUtil.isMethodSupported(np,
NamenodeProtocolPB.class, RPC.RpcKind.RPC_PROTOCOL_BUFFER,
RPC.getProtocolVersion(NamenodeProtocolPB.class), "rollEditLog");
assertTrue(exists);
exists = translator.isMethodSupported("bogusMethod");
exists = RpcClientUtil.isMethodSupported(np,
NamenodeProtocolPB.class, RPC.RpcKind.RPC_PROTOCOL_BUFFER,
RPC.getProtocolVersion(NamenodeProtocolPB.class), "bogusMethod");
assertFalse(exists);
}
@Test
public void testDatanodeProtocol() throws IOException {
DatanodeProtocolClientSideTranslatorPB translator =
@ -107,16 +115,18 @@ public void testClientDatanodeProtocol() throws IOException {
NetUtils.getDefaultSocketFactory(conf));
assertTrue(translator.isMethodSupported("refreshNamenodes"));
}
@Test
public void testClientNamenodeProtocol() throws IOException {
ClientNamenodeProtocolTranslatorPB translator =
(ClientNamenodeProtocolTranslatorPB) NameNodeProxies.createNonHAProxy(
ClientProtocol cp =
NameNodeProxies.createNonHAProxy(
conf, nnAddress, ClientProtocol.class,
UserGroupInformation.getCurrentUser(), true).getProxy();
assertTrue(translator.isMethodSupported("mkdirs"));
RpcClientUtil.isMethodSupported(cp,
ClientNamenodeProtocolPB.class, RPC.RpcKind.RPC_PROTOCOL_BUFFER,
RPC.getProtocolVersion(ClientNamenodeProtocolPB.class), "mkdirs");
}
@Test
public void tesJournalProtocol() throws IOException {
JournalProtocolTranslatorPB translator = (JournalProtocolTranslatorPB)