bharatviswa504 commented on a change in pull request #1006: HDDS-1723. Create new OzoneManagerLock class. URL: https://github.com/apache/hadoop/pull/1006#discussion_r297003664
########## File path: hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/lock/OzoneManagerLock.java ########## @@ -0,0 +1,336 @@ +/** + * 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.ozone.om.lock; + + +import java.util.ArrayList; +import java.util.List; + +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.ozone.lock.LockManager; + +/** + * Provides different locks to handle concurrency in OzoneMaster. + * We also maintain lock hierarchy, based on the weight. + * + * <table> + * <caption></caption> + * <tr> + * <td><b> WEIGHT </b></td> <td><b> LOCK </b></td> + * </tr> + * <tr> + * <td> 0 </td> <td> S3 Bucket Lock </td> + * </tr> + * <tr> + * <td> 1 </td> <td> Volume Lock </td> + * </tr> + * <tr> + * <td> 2 </td> <td> Bucket Lock </td> + * </tr> + * <tr> + * <td> 3 </td> <td> User Lock </td> + * </tr> + * <tr> + * <td> 4 </td> <td> S3 Secret Lock</td> + * </tr> + * <tr> + * <td> 5 </td> <td> Prefix Lock </td> + * </tr> + * </table> + * + * One cannot obtain a lower weight lock while holding a lock with higher + * weight. The other way around is possible. <br> + * <br> + * <p> + * For example: + * <br> + * {@literal ->} acquire volume lock (will work)<br> + * {@literal +->} acquire bucket lock (will work)<br> + * {@literal +-->} acquire s3 bucket lock (will throw Exception)<br> + * </p> + * <br> + */ + +public class OzoneManagerLock { + + private static final Logger LOG = + LoggerFactory.getLogger(OzoneManagerLock.class); + + private final LockManager<String> manager; + private final ThreadLocal<Short> lockSet = ThreadLocal.withInitial( + () -> Short.valueOf((short)0)); + + + /** + * Creates new OzoneManagerLock instance. + * @param conf Configuration object + */ + public OzoneManagerLock(Configuration conf) { + manager = new LockManager<>(conf); + } + + /** + * Acquire lock on resource. + * + * For S3_Bucket, VOLUME, BUCKET type resource, same thread acquiring lock + * again is allowed. + * + * For USER, PREFIX, S3_SECRET type resource, same thread acquiring lock + * again is not allowed. + * + * Special Note for UserLock: Single thread can acquire single user lock/ + * multi user lock. But not both at the same time. + * @param resourceName - Resource name on which user want to acquire lock. + * @param resource - Type of the resource. + */ + public void acquireLock(String resourceName, Resource resource) { + if (!resource.canLock(lockSet.get())) { + String errorMessage = getErrorMessage(resource); + LOG.error(errorMessage); + throw new RuntimeException(errorMessage); + } else { + manager.lock(resourceName); + lockSet.set(resource.setLock(lockSet.get())); + } + } + + private String getErrorMessage(Resource resource) { + return "Thread '" + Thread.currentThread().getName() + "' cannot " + + "acquire " + resource.name + " lock while holding " + + getCurrentLocks().toString() + " lock(s)."; + + } + + private List<String> getCurrentLocks() { + List<String> currentLocks = new ArrayList<>(); + int i=0; + short lockSetVal = lockSet.get(); + for (Resource value : Resource.values()) { + if (value.isLevelLocked(lockSetVal)) { + currentLocks.add(value.getName()); + } + } + return currentLocks; + } + + /** + * Acquire lock on multiple users. + * @param oldUserResource + * @param newUserResource + */ + public void acquireMultiUserLock(String oldUserResource, + String newUserResource) { + Resource resource = Resource.USER; + if (!resource.canLock(lockSet.get())) { + String errorMessage = getErrorMessage(resource); + LOG.error(errorMessage); + throw new RuntimeException(errorMessage); + } else { + // When acquiring multiple user locks, the reason for doing lexical + // order comparision is to avoid deadlock scenario. + + // Example: 1st thread acquire lock(ozone, hdfs) + // 2nd thread acquire lock(hdfs, ozone). + // If we don't acquire user locks in an order, there can be a deadlock. + + // 1st thread acquired lock on ozone, waiting for lock on hdfs, 2nd + // thread acquired lock on hdfs, waiting for lock on ozone. + // To avoid this when we acquire lock on multiple users, we acquire + // locks in lexical order, which can help us to avoid dead locks. + // Now if first thread acquires lock on hdfs, 2nd thread wait for lock + // on hdfs, and first thread acquires lock on ozone. Once after first + // thread releases user locks, 2nd thread acquires them. + + int compare = newUserResource.compareTo(oldUserResource); + String temp; + + // Order the user names in sorted order. Swap them. + if (compare > 0) { + temp = newUserResource; + newUserResource = oldUserResource; + oldUserResource = temp; + } + + if (compare == 0) { + // both users are equal. + manager.lock(oldUserResource); + } else { + manager.lock(newUserResource); + try { + manager.lock(oldUserResource); + } catch (Exception ex) { + // We got an exception acquiring 2nd user lock. Release already + // acquired user lock, and throw exception to the user. + manager.unlock(newUserResource); + throw ex; + } + } + lockSet.set(resource.setLock(lockSet.get())); + } + } + + + + /** + * Acquire lock on multiple users. + * @param oldUserResource + * @param newUserResource + */ + public void releaseMultiUserLock(String oldUserResource, + String newUserResource) { + Resource resource = Resource.USER; + int compare = newUserResource.compareTo(oldUserResource); + + String temp; + + // Order the user names in sorted order. Swap them. + if (compare > 0) { + temp = newUserResource; + newUserResource = oldUserResource; + oldUserResource = temp; + } + + if (compare == 0) { + // both users are equal. + manager.unlock(oldUserResource); + } else { + manager.unlock(newUserResource); + manager.unlock(oldUserResource); + } + lockSet.set(resource.clearLock(lockSet.get())); + } + + + public void releaseLock(String resourceName, Resource resource) { + + // TODO: Not checking release of higher order level lock happened while + // releasing lower order level lock, as for that we need counter for + // locks, as some locks support acquiring lock again. + manager.unlock(resourceName); + // clear lock + lockSet.set(resource.clearLock(lockSet.get())); + + } + + /** + * Resource defined in Ozone. + */ + public enum Resource { + // For S3 Bucket need to allow only for S3, that should be means only 1. + S3_BUCKET((byte) 0, "S3_BUCKET"), // = 1 + + // For volume need to allow both s3 bucket and volume. 01 + 10 = 11 (3) + VOLUME((byte) 1, "VOLUME"), // = 2 + + // For bucket we need to allow both s3 bucket, volume and bucket. Which + // is equal to 100 + 010 + 001 = 111 = 4 + 2 + 1 = 7 + BUCKET((byte) 2, "BUCKET"), // = 4 + + // For user we need to allow s3 bucket, volume, bucket and user lock. + // Which is 8 4 + 2 + 1 = 15 + USER((byte) 3, "USER"), // 15 + + S3_SECRET((byte) 4, "S3_SECRET"), // 31 + PREFIX((byte) 5, "PREFIX"); //63 + + // level of the resource + private byte lockLevel; + + // This will tell the value, till which we can allow locking. + private short mask; + + // This value will help during setLock, and also will tell whether we can + // re-acquire lock or not. + private short setMask; + + // Name of the resource. + private String name; + + Resource(byte pos, String name) { + this.lockLevel = pos; + this.mask += (short) Math.pow(2, lockLevel + 1) - 1; + this.setMask = (short) Math.pow(2, lockLevel); + this.name = name; + } + + boolean canLock(short lockSetVal) { + + // For USER, S3_SECRET and PREFIX we shall not allow re-acquire locks at + // from single thread. 2nd condition is we have acquired one of these + // locks, but after that trying to acquire a lock with less than equal of + // lockLevel, we should disallow. + if (((USER.setMask & lockSetVal) == USER.setMask || + (S3_SECRET.setMask & lockSetVal) == S3_SECRET.setMask || + (PREFIX.setMask & lockSetVal) == PREFIX.setMask) + && setMask <= lockSetVal) { + return false; + } + + + // Our mask is the summation of bits of all previous possible locks. In + // other words it is the largest possible value for that bit position. + + // For example for Volume lock, bit position is 1, and mask is 3. Which + // is the largest value that can be represented with 2 bits is 3. + // Therefore if lockSet is larger than mask we have to return false i.e + // some other higher order lock has been acquired. + + return lockSetVal <= mask; + } + + /** + * Set Lock bits in lockSetVal. + * + * @param lockSetVal + * @return Updated value which has set lock bits. + */ + short setLock(short lockSetVal) { + LOG.debug("acquire" + name + (short) (lockSetVal | setMask)); + return (short) (lockSetVal | setMask); + } + + /** + * Clear lock from lockSetVal. + * + * @param lockSetVal + * @return Updated value which has cleared lock bits. + */ + short clearLock(short lockSetVal) { + LOG.debug("release" + name + (short) (lockSetVal & ~setMask)); Review comment: Fixed as suggested. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
