Revert "HDFS-10346. Implement asynchronous setPermission/setOwner for DistributedFileSystem. Contributed by Xiaobing Zhou"

This reverts commit 7251bb922b.
This commit is contained in:
Andrew Wang 2016-06-03 18:09:17 -07:00
parent e4450d47f1
commit f23d5dfc60
4 changed files with 43 additions and 351 deletions

View File

@ -27,7 +27,6 @@
import org.apache.hadoop.classification.InterfaceStability.Unstable;
import org.apache.hadoop.fs.Options;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.permission.FsPermission;
import org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB;
import org.apache.hadoop.ipc.Client;
@ -38,9 +37,6 @@
* This instance of this class is the way end-user code interacts
* with a Hadoop DistributedFileSystem in an asynchronous manner.
*
* This class is unstable, so no guarantee is provided as to reliability,
* stability or compatibility across any level of release granularity.
*
*****************************************************************/
@Unstable
public class AsyncDistributedFileSystem {
@ -115,59 +111,4 @@ public Future<Void> rename(Path src, Path dst,
Client.setAsynchronousMode(isAsync);
}
}
/**
* Set permission of a path.
*
* @param p
* the path the permission is set to
* @param permission
* the permission that is set to a path.
* @return an instance of Future, #get of which is invoked to wait for
* asynchronous call being finished.
*/
public Future<Void> setPermission(Path p, final FsPermission permission)
throws IOException {
dfs.getFsStatistics().incrementWriteOps(1);
final Path absPath = dfs.fixRelativePart(p);
final boolean isAsync = Client.isAsynchronousMode();
Client.setAsynchronousMode(true);
try {
dfs.getClient().setPermission(dfs.getPathName(absPath), permission);
return getReturnValue();
} finally {
Client.setAsynchronousMode(isAsync);
}
}
/**
* Set owner of a path (i.e. a file or a directory). The parameters username
* and groupname cannot both be null.
*
* @param p
* The path
* @param username
* If it is null, the original username remains unchanged.
* @param groupname
* If it is null, the original groupname remains unchanged.
* @return an instance of Future, #get of which is invoked to wait for
* asynchronous call being finished.
*/
public Future<Void> setOwner(Path p, String username, String groupname)
throws IOException {
if (username == null && groupname == null) {
throw new IOException("username == null && groupname == null");
}
dfs.getFsStatistics().incrementWriteOps(1);
final Path absPath = dfs.fixRelativePart(p);
final boolean isAsync = Client.isAsynchronousMode();
Client.setAsynchronousMode(true);
try {
dfs.getClient().setOwner(dfs.getPathName(absPath), username, groupname);
return getReturnValue();
} finally {
Client.setAsynchronousMode(isAsync);
}
}
}

View File

@ -367,30 +367,12 @@ public void setPermission(String src, FsPermission permission)
.setPermission(PBHelperClient.convert(permission))
.build();
try {
if (Client.isAsynchronousMode()) {
rpcProxy.setPermission(null, req);
setReturnValueCallback();
} else {
rpcProxy.setPermission(null, req);
}
rpcProxy.setPermission(null, req);
} catch (ServiceException e) {
throw ProtobufHelper.getRemoteException(e);
}
}
private void setReturnValueCallback() {
final Callable<Message> returnMessageCallback = ProtobufRpcEngine
.getReturnMessageCallback();
Callable<Void> callBack = new Callable<Void>() {
@Override
public Void call() throws Exception {
returnMessageCallback.call();
return null;
}
};
RETURN_VALUE_CALLBACK.set(callBack);
}
@Override
public void setOwner(String src, String username, String groupname)
throws IOException {
@ -401,12 +383,7 @@ public void setOwner(String src, String username, String groupname)
if (groupname != null)
req.setGroupname(groupname);
try {
if (Client.isAsynchronousMode()) {
rpcProxy.setOwner(null, req.build());
setReturnValueCallback();
} else {
rpcProxy.setOwner(null, req.build());
}
rpcProxy.setOwner(null, req.build());
} catch (ServiceException e) {
throw ProtobufHelper.getRemoteException(e);
}
@ -536,7 +513,17 @@ public void rename2(String src, String dst, Rename... options)
try {
if (Client.isAsynchronousMode()) {
rpcProxy.rename2(null, req);
setReturnValueCallback();
final Callable<Message> returnMessageCallback = ProtobufRpcEngine
.getReturnMessageCallback();
Callable<Void> callBack = new Callable<Void>() {
@Override
public Void call() throws Exception {
returnMessageCallback.call();
return null;
}
};
RETURN_VALUE_CALLBACK.set(callBack);
} else {
rpcProxy.rename2(null, req);
}

View File

@ -22,11 +22,8 @@
import java.io.IOException;
import java.security.PrivilegedExceptionAction;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Random;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.Future;
@ -34,30 +31,18 @@
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.CommonConfigurationKeys;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.Options.Rename;
import org.apache.hadoop.fs.permission.FsAction;
import org.apache.hadoop.fs.permission.FsPermission;
import org.apache.hadoop.hdfs.TestDFSPermission.PermissionGenerator;
import org.apache.hadoop.hdfs.protocol.LocatedBlocks;
import org.apache.hadoop.hdfs.server.blockmanagement.BlockManager;
import org.apache.hadoop.hdfs.server.namenode.NameNodeAdapter;
import org.apache.hadoop.ipc.AsyncCallLimitExceededException;
import org.apache.hadoop.security.AccessControlException;
import org.apache.hadoop.security.UserGroupInformation;
import org.apache.hadoop.util.Time;
import org.junit.Test;
public class TestAsyncDFSRename {
public static final Log LOG = LogFactory.getLog(TestAsyncDFSRename.class);
private final long seed = Time.now();
private final Random r = new Random(seed);
private final PermissionGenerator permGenerator = new PermissionGenerator(r);
private final short replFactor = 2;
private final long blockSize = 512;
private long fileLen = blockSize * 3;
/**
* Check the blocks of dst file are cleaned after rename with overwrite
@ -65,6 +50,8 @@ public class TestAsyncDFSRename {
*/
@Test(timeout = 60000)
public void testAsyncRenameWithOverwrite() throws Exception {
final short replFactor = 2;
final long blockSize = 512;
Configuration conf = new Configuration();
MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).numDataNodes(
replFactor).build();
@ -73,6 +60,8 @@ public void testAsyncRenameWithOverwrite() throws Exception {
AsyncDistributedFileSystem adfs = dfs.getAsyncDistributedFileSystem();
try {
long fileLen = blockSize * 3;
String src = "/foo/src";
String dst = "/foo/dst";
String src2 = "/foo/src2";
@ -126,6 +115,8 @@ public void testAsyncRenameWithOverwrite() throws Exception {
@Test(timeout = 60000)
public void testCallGetReturnValueMultipleTimes() throws Exception {
final short replFactor = 2;
final long blockSize = 512;
final Path renameDir = new Path(
"/test/testCallGetReturnValueMultipleTimes/");
final Configuration conf = new HdfsConfiguration();
@ -136,6 +127,7 @@ public void testCallGetReturnValueMultipleTimes() throws Exception {
final DistributedFileSystem dfs = cluster.getFileSystem();
final AsyncDistributedFileSystem adfs = dfs.getAsyncDistributedFileSystem();
final int count = 100;
long fileLen = blockSize * 3;
final Map<Integer, Future<Void>> returnFutures = new HashMap<Integer, Future<Void>>();
assertTrue(dfs.mkdirs(renameDir));
@ -186,15 +178,15 @@ private void verifyCallGetReturnValueMultipleTimes(
}
}
@Test
public void testConservativeConcurrentAsyncRenameWithOverwrite()
@Test(timeout = 120000)
public void testAggressiveConcurrentAsyncRenameWithOverwrite()
throws Exception {
internalTestConcurrentAsyncRenameWithOverwrite(100,
"testAggressiveConcurrentAsyncRenameWithOverwrite");
}
@Test(timeout = 60000)
public void testAggressiveConcurrentAsyncRenameWithOverwrite()
public void testConservativeConcurrentAsyncRenameWithOverwrite()
throws Exception {
internalTestConcurrentAsyncRenameWithOverwrite(10000,
"testConservativeConcurrentAsyncRenameWithOverwrite");
@ -202,6 +194,8 @@ public void testAggressiveConcurrentAsyncRenameWithOverwrite()
private void internalTestConcurrentAsyncRenameWithOverwrite(
final int asyncCallLimit, final String basePath) throws Exception {
final short replFactor = 2;
final long blockSize = 512;
final Path renameDir = new Path(String.format("/test/%s/", basePath));
Configuration conf = new HdfsConfiguration();
conf.setInt(CommonConfigurationKeys.IPC_CLIENT_ASYNC_CALLS_MAX_KEY,
@ -212,6 +206,7 @@ private void internalTestConcurrentAsyncRenameWithOverwrite(
DistributedFileSystem dfs = cluster.getFileSystem();
AsyncDistributedFileSystem adfs = dfs.getAsyncDistributedFileSystem();
int count = 1000;
long fileLen = blockSize * 3;
int start = 0, end = 0;
Map<Integer, Future<Void>> returnFutures = new HashMap<Integer, Future<Void>>();
@ -279,206 +274,8 @@ private void waitForReturnValues(
}
}
@Test
public void testConservativeConcurrentAsyncAPI() throws Exception {
internalTestConcurrentAsyncAPI(100, "testConservativeConcurrentAsyncAPI");
}
@Test(timeout = 60000)
public void testAggressiveConcurrentAsyncAPI() throws Exception {
internalTestConcurrentAsyncAPI(10000, "testAggressiveConcurrentAsyncAPI");
}
private void internalTestConcurrentAsyncAPI(final int asyncCallLimit,
final String basePath) throws Exception {
Configuration conf = new HdfsConfiguration();
String group1 = "group1";
String group2 = "group2";
String user1 = "user1";
int count = 500;
// explicitly turn on permission checking
conf.setBoolean(DFSConfigKeys.DFS_PERMISSIONS_ENABLED_KEY, true);
// set the limit of max async calls
conf.setInt(CommonConfigurationKeys.IPC_CLIENT_ASYNC_CALLS_MAX_KEY,
asyncCallLimit);
// create fake mapping for the groups
Map<String, String[]> u2gMap = new HashMap<String, String[]>(1);
u2gMap.put(user1, new String[] {group1, group2});
DFSTestUtil.updateConfWithFakeGroupMapping(conf, u2gMap);
// start mini cluster
final MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf)
.numDataNodes(3).build();
cluster.waitActive();
AsyncDistributedFileSystem adfs = cluster.getFileSystem()
.getAsyncDistributedFileSystem();
// prepare for test
FileSystem rootFs = FileSystem.get(conf);
final Path parent = new Path(String.format("/test/%s/", basePath));
final Path[] srcs = new Path[count];
final Path[] dsts = new Path[count];
short[] permissions = new short[count];
for (int i = 0; i < count; i++) {
srcs[i] = new Path(parent, "src" + i);
dsts[i] = new Path(parent, "dst" + i);
DFSTestUtil.createFile(rootFs, srcs[i], fileLen, replFactor, 1);
DFSTestUtil.createFile(rootFs, dsts[i], fileLen, replFactor, 1);
assertTrue(rootFs.exists(srcs[i]));
assertTrue(rootFs.getFileStatus(srcs[i]).isFile());
assertTrue(rootFs.exists(dsts[i]));
assertTrue(rootFs.getFileStatus(dsts[i]).isFile());
permissions[i] = permGenerator.next();
}
Map<Integer, Future<Void>> renameRetFutures =
new HashMap<Integer, Future<Void>>();
Map<Integer, Future<Void>> permRetFutures =
new HashMap<Integer, Future<Void>>();
Map<Integer, Future<Void>> ownerRetFutures =
new HashMap<Integer, Future<Void>>();
int start = 0, end = 0;
// test rename
for (int i = 0; i < count; i++) {
for (;;) {
try {
Future<Void> returnFuture = adfs.rename(srcs[i], dsts[i],
Rename.OVERWRITE);
renameRetFutures.put(i, returnFuture);
break;
} catch (AsyncCallLimitExceededException e) {
start = end;
end = i;
waitForReturnValues(renameRetFutures, start, end);
}
}
}
// wait for completing the calls
for (int i = start; i < count; i++) {
renameRetFutures.get(i).get();
}
// Restart NN and check the rename successfully
cluster.restartNameNodes();
// very the src should not exist, dst should
for (int i = 0; i < count; i++) {
assertFalse(rootFs.exists(srcs[i]));
assertTrue(rootFs.exists(dsts[i]));
}
// test permissions
try {
for (int i = 0; i < count; i++) {
for (;;) {
try {
Future<Void> retFuture = adfs.setPermission(dsts[i],
new FsPermission(permissions[i]));
permRetFutures.put(i, retFuture);
break;
} catch (AsyncCallLimitExceededException e) {
start = end;
end = i;
waitForReturnValues(permRetFutures, start, end);
}
}
}
// wait for completing the calls
for (int i = start; i < count; i++) {
permRetFutures.get(i).get();
}
// Restart NN and check permission then
cluster.restartNameNodes();
// verify the permission
for (int i = 0; i < count; i++) {
assertTrue(rootFs.exists(dsts[i]));
FsPermission fsPerm = new FsPermission(permissions[i]);
checkAccessPermissions(rootFs.getFileStatus(dsts[i]),
fsPerm.getUserAction());
}
// test setOwner
start = 0;
end = 0;
for (int i = 0; i < count; i++) {
for (;;) {
try {
Future<Void> retFuture = adfs.setOwner(dsts[i], "user1",
"group2");
ownerRetFutures.put(i, retFuture);
break;
} catch (AsyncCallLimitExceededException e) {
start = end;
end = i;
waitForReturnValues(ownerRetFutures, start, end);
}
}
}
// wait for completing the calls
for (int i = start; i < count; i++) {
ownerRetFutures.get(i).get();
}
// Restart NN and check owner then
cluster.restartNameNodes();
// verify the owner
for (int i = 0; i < count; i++) {
assertTrue(rootFs.exists(dsts[i]));
assertTrue(
"user1".equals(rootFs.getFileStatus(dsts[i]).getOwner()));
assertTrue(
"group2".equals(rootFs.getFileStatus(dsts[i]).getGroup()));
}
} catch (AccessControlException ace) {
throw ace;
} finally {
if (rootFs != null) {
rootFs.close();
}
if (cluster != null) {
cluster.shutdown();
}
}
}
static void checkAccessPermissions(FileStatus stat, FsAction mode)
throws IOException {
checkAccessPermissions(UserGroupInformation.getCurrentUser(), stat, mode);
}
static void checkAccessPermissions(final UserGroupInformation ugi,
FileStatus stat, FsAction mode) throws IOException {
FsPermission perm = stat.getPermission();
String user = ugi.getShortUserName();
List<String> groups = Arrays.asList(ugi.getGroupNames());
if (user.equals(stat.getOwner())) {
if (perm.getUserAction().implies(mode)) {
return;
}
} else if (groups.contains(stat.getGroup())) {
if (perm.getGroupAction().implies(mode)) {
return;
}
} else {
if (perm.getOtherAction().implies(mode)) {
return;
}
}
throw new AccessControlException(String.format(
"Permission denied: user=%s, path=\"%s\":%s:%s:%s%s", user, stat
.getPath(), stat.getOwner(), stat.getGroup(),
stat.isDirectory() ? "d" : "-", perm));
}
@Test(timeout = 60000)
public void testAsyncAPIWithException() throws Exception {
public void testAsyncRenameWithException() throws Exception {
Configuration conf = new HdfsConfiguration();
String group1 = "group1";
String group2 = "group2";
@ -489,9 +286,9 @@ public void testAsyncAPIWithException() throws Exception {
conf.setBoolean(DFSConfigKeys.DFS_PERMISSIONS_ENABLED_KEY, true);
// create fake mapping for the groups
Map<String, String[]> u2gMap = new HashMap<String, String[]>(1);
u2gMap.put(user1, new String[] {group1, group2});
DFSTestUtil.updateConfWithFakeGroupMapping(conf, u2gMap);
Map<String, String[]> u2g_map = new HashMap<String, String[]>(1);
u2g_map.put(user1, new String[] { group1, group2 });
DFSTestUtil.updateConfWithFakeGroupMapping(conf, u2g_map);
// Initiate all four users
ugi1 = UserGroupInformation.createUserForTesting(user1, new String[] {
@ -502,7 +299,7 @@ public void testAsyncAPIWithException() throws Exception {
cluster.waitActive();
FileSystem rootFs = FileSystem.get(conf);
final Path renameDir = new Path("/test/async_api_exception/");
final Path renameDir = new Path("/test/async_rename_exception/");
final Path src = new Path(renameDir, "src");
final Path dst = new Path(renameDir, "dst");
rootFs.mkdirs(src);
@ -515,33 +312,11 @@ public AsyncDistributedFileSystem run() throws Exception {
}
});
Future<Void> retFuture;
try {
retFuture = adfs.rename(src, dst, Rename.OVERWRITE);
retFuture.get();
Future<Void> returnFuture = adfs.rename(src, dst, Rename.OVERWRITE);
returnFuture.get();
} catch (ExecutionException e) {
checkPermissionDenied(e, src, user1);
assertTrue("Permission denied messages must carry the path parent", e
.getMessage().contains(src.getParent().toUri().getPath()));
}
FsPermission fsPerm = new FsPermission(permGenerator.next());
try {
retFuture = adfs.setPermission(src, fsPerm);
retFuture.get();
} catch (ExecutionException e) {
checkPermissionDenied(e, src, user1);
assertTrue("Permission denied messages must carry the name of the path",
e.getMessage().contains(src.getName()));
}
try {
retFuture = adfs.setOwner(src, "user1", "group2");
retFuture.get();
} catch (ExecutionException e) {
checkPermissionDenied(e, src, user1);
assertTrue("Permission denied messages must carry the name of the path",
e.getMessage().contains(src.getName()));
} finally {
if (rootFs != null) {
rootFs.close();
@ -559,5 +334,7 @@ private void checkPermissionDenied(final Exception e, final Path dir,
e.getMessage().contains("AccessControlException"));
assertTrue("Permission denied messages must carry the username", e
.getMessage().contains(user));
assertTrue("Permission denied messages must carry the path parent", e
.getMessage().contains(dir.getParent().toUri().getPath()));
}
}

View File

@ -196,35 +196,22 @@ private short getPermission(Path path) throws IOException {
return fs.getFileStatus(path).getPermission().toShort();
}
private void create(OpType op, Path name) throws IOException {
create(fs, conf, op, name);
}
/* create a file/directory with the default umask and permission */
static void create(final FileSystem fs, final Configuration fsConf,
OpType op, Path name) throws IOException {
create(fs, fsConf, op, name, DEFAULT_UMASK, new FsPermission(
DEFAULT_PERMISSION));
}
private void create(OpType op, Path name, short umask,
FsPermission permission)
throws IOException {
create(fs, conf, op, name, umask, permission);
private void create(OpType op, Path name) throws IOException {
create(op, name, DEFAULT_UMASK, new FsPermission(DEFAULT_PERMISSION));
}
/* create a file/directory with the given umask and permission */
static void create(final FileSystem fs, final Configuration fsConf,
OpType op, Path name, short umask, FsPermission permission)
throws IOException {
private void create(OpType op, Path name, short umask,
FsPermission permission) throws IOException {
// set umask in configuration, converting to padded octal
fsConf.set(FsPermission.UMASK_LABEL, String.format("%1$03o", umask));
conf.set(FsPermission.UMASK_LABEL, String.format("%1$03o", umask));
// create the file/directory
switch (op) {
case CREATE:
FSDataOutputStream out = fs.create(name, permission, true,
fsConf.getInt(CommonConfigurationKeys.IO_FILE_BUFFER_SIZE_KEY, 4096),
conf.getInt(CommonConfigurationKeys.IO_FILE_BUFFER_SIZE_KEY, 4096),
fs.getDefaultReplication(name), fs.getDefaultBlockSize(name), null);
out.close();
break;
@ -372,7 +359,7 @@ private void checkOwnership(Path name, String expectedOwner,
final static private String DIR_NAME = "dir";
final static private String FILE_DIR_NAME = "filedir";
enum OpType {CREATE, MKDIRS, OPEN, SET_REPLICATION,
private enum OpType {CREATE, MKDIRS, OPEN, SET_REPLICATION,
GET_FILEINFO, IS_DIR, EXISTS, GET_CONTENT_LENGTH, LIST, RENAME, DELETE
};
@ -628,7 +615,7 @@ private void testPermissionCheckingPerUser(UserGroupInformation ugi,
/* A random permission generator that guarantees that each permission
* value is generated only once.
*/
static class PermissionGenerator {
static private class PermissionGenerator {
private final Random r;
private final short[] permissions = new short[MAX_PERMISSION + 1];
private int numLeft = MAX_PERMISSION + 1;