[
https://issues.apache.org/jira/browse/HDFS-17384?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17851031#comment-17851031
]
ASF GitHub Bot commented on HDFS-17384:
---------------------------------------
zhengchenyu commented on code in PR #6762:
URL: https://github.com/apache/hadoop/pull/6762#discussion_r1622166340
##########
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/fgl/FineGrainedFSNamesystemLock.java:
##########
@@ -0,0 +1,300 @@
+/**
+ * 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.namenode.fgl;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hdfs.server.namenode.FSNamesystemLock;
+import org.apache.hadoop.metrics2.lib.MutableRatesWithAggregation;
+
+import java.util.concurrent.locks.ReentrantReadWriteLock;
+import java.util.function.Supplier;
+
+/**
+ * Splitting the global FSN lock into FSLock and BMLock.
+ * FSLock is used to protect directory tree-related operations.
+ * BMLock is used to protect block-related and dn-related operations.
+ * The lock order should be: FSLock,BMLock.
+ */
+public class FineGrainedFSNamesystemLock implements FSNLockManager {
+ private final FSNamesystemLock fsLock;
+ private final FSNamesystemLock bmLock;
+
+ public FineGrainedFSNamesystemLock(Configuration conf,
MutableRatesWithAggregation aggregation) {
+ this.fsLock = new FSNamesystemLock(conf, "FS", aggregation);
+ this.bmLock = new FSNamesystemLock(conf, "BM", aggregation);
+ }
+
+ @Override
+ public void readLock(FSNamesystemLockMode lockMode) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.fsLock.readLock();
+ this.bmLock.readLock();
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.readLock();
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.readLock();
+ }
+ }
+
+ public void readLockInterruptibly(FSNamesystemLockMode lockMode) throws
InterruptedException {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.fsLock.readLockInterruptibly();
+ try {
+ this.bmLock.readLockInterruptibly();
+ } catch (InterruptedException e) {
+ // The held FSLock should be released if the current thread is
interrupted
+ // while acquiring the BMLock.
+ this.fsLock.readUnlock("BMReadLockInterruptiblyFailed");
+ throw e;
+ }
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.readLockInterruptibly();
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.readLockInterruptibly();
+ }
+ }
+
+ @Override
+ public void readUnlock(FSNamesystemLockMode lockMode, String opName) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.bmLock.readUnlock(opName);
+ this.fsLock.readUnlock(opName);
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.readUnlock(opName);
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.readUnlock(opName);
+ }
+ }
+
+ public void readUnlock(FSNamesystemLockMode lockMode, String opName,
+ Supplier<String> lockReportInfoSupplier) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.bmLock.readUnlock(opName, lockReportInfoSupplier);
+ this.fsLock.readUnlock(opName, lockReportInfoSupplier);
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.readUnlock(opName, lockReportInfoSupplier);
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.readUnlock(opName, lockReportInfoSupplier);
+ }
+ }
+
+ @Override
+ public void writeLock(FSNamesystemLockMode lockMode) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.fsLock.writeLock();
+ this.bmLock.writeLock();
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.writeLock();
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.writeLock();
+ }
+ }
+
+ @Override
+ public void writeUnlock(FSNamesystemLockMode lockMode, String opName) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.bmLock.writeUnlock(opName);
+ this.fsLock.writeUnlock(opName);
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.writeUnlock(opName);
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.writeUnlock(opName);
+ }
+ }
+
+ @Override
+ public void writeUnlock(FSNamesystemLockMode lockMode, String opName,
+ boolean suppressWriteLockReport) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.bmLock.writeUnlock(opName, suppressWriteLockReport);
+ this.fsLock.writeUnlock(opName, suppressWriteLockReport);
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.writeUnlock(opName, suppressWriteLockReport);
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.writeUnlock(opName, suppressWriteLockReport);
+ }
+ }
+
+ public void writeUnlock(FSNamesystemLockMode lockMode, String opName,
+ Supplier<String> lockReportInfoSupplier) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.bmLock.writeUnlock(opName, lockReportInfoSupplier);
+ this.fsLock.writeUnlock(opName, lockReportInfoSupplier);
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.writeUnlock(opName, lockReportInfoSupplier);
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.writeUnlock(opName, lockReportInfoSupplier);
+ }
+ }
+
+ @Override
+ public void writeLockInterruptibly(FSNamesystemLockMode lockMode)
+ throws InterruptedException {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ this.fsLock.writeLockInterruptibly();
+ try {
+ this.bmLock.writeLockInterruptibly();
+ } catch (InterruptedException e) {
+ // The held FSLock should be released if the current thread is
interrupted
+ // while acquiring the BMLock.
+ this.fsLock.writeUnlock("BMWriteLockInterruptiblyFailed");
+ throw e;
+ }
+ } else if (lockMode.equals(FSNamesystemLockMode.FS)) {
+ this.fsLock.writeLockInterruptibly();
+ } else if (lockMode.equals(FSNamesystemLockMode.BM)) {
+ this.bmLock.writeLockInterruptibly();
+ }
+ }
+
+ @Override
+ public boolean hasWriteLock(FSNamesystemLockMode lockMode) {
+ if (lockMode.equals(FSNamesystemLockMode.GLOBAL)) {
+ if (this.fsLock.isWriteLockedByCurrentThread()) {
+ // The bm writeLock should be held by the current thread.
+ assert this.bmLock.isWriteLockedByCurrentThread();
Review Comment:
I found that hasWriteLock has two way to used:
* (1) Make sure we hold a lock. like `assert
fsn.hasReadLock(FSNamesystemLockMode.GLOBAL);`
* (2) Check whether there is a lock to determine whether to perform some
action. like `boolean hadFsnWriteLock =
fsn.hasWriteLock(FSNamesystemLockMode.GLOBAL);`
Here add `assert`, may throw AssertionError. So I think the new
`hasWriteLock` only have the semantics described in (1).
I think we need to separate two approaches to handle (1) and (2).
> [FGL] Replace the global lock with global FS Lock and global BM lock
> --------------------------------------------------------------------
>
> Key: HDFS-17384
> URL: https://issues.apache.org/jira/browse/HDFS-17384
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: ZanderXu
> Assignee: ZanderXu
> Priority: Major
> Labels: FGL, pull-request-available
>
> First, we can replace the current global lock with two locks, global FS lock
> and global BM lock.
> The global FS lock is used to make directory tree-related operations
> thread-safe.
> The global BM lock is used to make block-related operations and DN-related
> operations thread-safe.
>
> For some operations involving both directory tree and block or DN, the global
> FS lock and the global BM lock are acquired.
>
> The lock order should be:
> * The global FS lock
> * The global BM lock
>
> There are some special requirements for this ticket.
> * End-user can choose to use global lock or fine-grained lock through
> configuration.
> * Try not to modify the current implementation logic as much as possible.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
