aswinshakil commented on code in PR #5579: URL: https://github.com/apache/ozone/pull/5579#discussion_r1427123683
########## hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/service/SnapshotDirectoryCleaningService.java: ########## @@ -0,0 +1,520 @@ +/* + * 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 + * <p> + * http://www.apache.org/licenses/LICENSE-2.0 + * <p> + * 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.service; + +import com.google.common.annotations.VisibleForTesting; +import com.google.protobuf.ServiceException; +import org.apache.commons.lang3.StringUtils; +import org.apache.hadoop.hdds.client.BlockID; +import org.apache.hadoop.hdds.scm.protocol.ScmBlockLocationProtocol; +import org.apache.hadoop.hdds.utils.BackgroundTask; +import org.apache.hadoop.hdds.utils.BackgroundTaskQueue; +import org.apache.hadoop.hdds.utils.BackgroundTaskResult; +import org.apache.hadoop.hdds.utils.db.Table; +import org.apache.hadoop.hdds.utils.db.TableIterator; +import org.apache.hadoop.ozone.common.BlockGroup; +import org.apache.hadoop.ozone.om.IOmMetadataReader; +import org.apache.hadoop.ozone.om.OMMetadataManager; +import org.apache.hadoop.ozone.om.OmMetadataManagerImpl; +import org.apache.hadoop.ozone.om.OmSnapshot; +import org.apache.hadoop.ozone.om.OmSnapshotManager; +import org.apache.hadoop.ozone.om.OzoneManager; +import org.apache.hadoop.ozone.om.SnapshotChainManager; +import org.apache.hadoop.ozone.om.helpers.OMRatisHelper; +import org.apache.hadoop.ozone.om.helpers.OmBucketInfo; +import org.apache.hadoop.ozone.om.helpers.OmDirectoryInfo; +import org.apache.hadoop.ozone.om.helpers.OmKeyInfo; +import org.apache.hadoop.ozone.om.helpers.OmKeyLocationInfoGroup; +import org.apache.hadoop.ozone.om.helpers.SnapshotInfo; +import org.apache.hadoop.ozone.om.ratis.OzoneManagerRatisServer; +import org.apache.hadoop.ozone.om.request.file.OMFileRequest; +import org.apache.hadoop.ozone.om.snapshot.ReferenceCounted; +import org.apache.hadoop.ozone.om.snapshot.SnapshotCache; +import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.OMRequest; +import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.SetSnapshotPropertyRequest; +import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.SnapshotSize; +import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.Type; +import org.apache.ratis.protocol.ClientId; +import org.apache.ratis.protocol.Message; +import org.apache.ratis.protocol.RaftClientRequest; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Stack; +import java.util.concurrent.TimeUnit; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.stream.Collectors; + +import static org.apache.hadoop.ozone.om.OmSnapshotManager.getSnapshotPrefix; +import static org.apache.hadoop.ozone.om.helpers.SnapshotInfo.SnapshotStatus.SNAPSHOT_ACTIVE; +import static org.apache.hadoop.ozone.om.request.file.OMFileRequest.getDirectoryInfo; +import static org.apache.hadoop.ozone.om.snapshot.SnapshotUtils.getOzonePathKeyForFso; + +/** + * Snapshot BG Service for deleted directory deep clean and exclusive size + * calculation for deleted directories. + */ +public class SnapshotDirectoryCleaningService + extends AbstractKeyDeletingService { + // Use only a single thread for DirDeletion. Multiple threads would read + // or write to same tables and can send deletion requests for same key + // multiple times. + private static final int SNAPSHOT_DIR_CORE_POOL_SIZE = 1; + + private final AtomicBoolean suspended; + private final Map<String, Long> exclusiveSizeMap; + private final Map<String, Long> exclusiveReplicatedSizeMap; + + public SnapshotDirectoryCleaningService(long interval, TimeUnit unit, + long serviceTimeout, + OzoneManager ozoneManager, + ScmBlockLocationProtocol scmClient) { + super(SnapshotDirectoryCleaningService.class.getSimpleName(), + interval, unit, SNAPSHOT_DIR_CORE_POOL_SIZE, serviceTimeout, + ozoneManager, scmClient); + this.suspended = new AtomicBoolean(false); + this.exclusiveSizeMap = new HashMap<>(); + this.exclusiveReplicatedSizeMap = new HashMap<>(); + } + + private boolean shouldRun() { + if (getOzoneManager() == null) { + // OzoneManager can be null for testing + return true; + } + return getOzoneManager().isLeaderReady() && !suspended.get(); + } + + /** + * Suspend the service. + */ + @VisibleForTesting + public void suspend() { + suspended.set(true); + } + + /** + * Resume the service if suspended. + */ + @VisibleForTesting + public void resume() { + suspended.set(false); + } + + @Override + public BackgroundTaskQueue getTasks() { + BackgroundTaskQueue queue = new BackgroundTaskQueue(); + queue.add(new SnapshotDirectoryCleaningService.SnapshotDirTask()); + return queue; + } + + private class SnapshotDirTask implements BackgroundTask { + + @Override + public BackgroundTaskResult call() { + if (!shouldRun()) { + return BackgroundTaskResult.EmptyTaskResult.newResult(); + } + LOG.debug("Running SnapshotDirectoryCleaningService"); + + getRunCount().incrementAndGet(); + OmSnapshotManager omSnapshotManager = + getOzoneManager().getOmSnapshotManager(); + Table<String, SnapshotInfo> snapshotInfoTable = + getOzoneManager().getMetadataManager().getSnapshotInfoTable(); + OmMetadataManagerImpl metadataManager = (OmMetadataManagerImpl) + getOzoneManager().getMetadataManager(); + SnapshotChainManager snapChainManager = metadataManager + .getSnapshotChainManager(); + + try (TableIterator<String, ? extends Table.KeyValue + <String, SnapshotInfo>> iterator = snapshotInfoTable.iterator()) { + + while (iterator.hasNext()) { + SnapshotInfo currSnapInfo = iterator.next().getValue(); + + // Expand deleted dirs only on active snapshot. Deleted Snapshots + // will be cleaned up by SnapshotDeletingService. + if (currSnapInfo.getSnapshotStatus() != SNAPSHOT_ACTIVE || + currSnapInfo.getDeepCleanedDeletedDir()) { + continue; + } + + long volumeId = metadataManager + .getVolumeId(currSnapInfo.getVolumeName()); + // Get bucketInfo for the snapshot bucket to get bucket layout. + String dbBucketKey = metadataManager + .getBucketKey(currSnapInfo.getVolumeName(), + currSnapInfo.getBucketName()); + OmBucketInfo bucketInfo = metadataManager + .getBucketTable().get(dbBucketKey); + + if (bucketInfo == null) { + throw new IllegalStateException("Bucket " + "/" + + currSnapInfo.getVolumeName() + "/" + currSnapInfo + .getBucketName() + " is not found. BucketInfo should not be " + + "null for snapshotted bucket. The OM is in unexpected state."); + } + + SnapshotInfo previousSnapshot = getPreviousActiveSnapshot( + currSnapInfo, snapChainManager, omSnapshotManager); + SnapshotInfo previousToPrevSnapshot = null; + + Table<String, OmKeyInfo> previousKeyTable = null; + Table<String, String> prevRenamedTable = null; + ReferenceCounted<IOmMetadataReader, SnapshotCache> + rcPrevOmSnapshot = null; + + if (previousSnapshot != null) { + rcPrevOmSnapshot = omSnapshotManager.checkForSnapshot( + previousSnapshot.getVolumeName(), + previousSnapshot.getBucketName(), + getSnapshotPrefix(previousSnapshot.getName()), false); + OmSnapshot omPreviousSnapshot = (OmSnapshot) + rcPrevOmSnapshot.get(); + + previousKeyTable = omPreviousSnapshot.getMetadataManager() + .getKeyTable(bucketInfo.getBucketLayout()); + prevRenamedTable = omPreviousSnapshot + .getMetadataManager().getSnapshotRenamedTable(); + previousToPrevSnapshot = getPreviousActiveSnapshot( + previousSnapshot, snapChainManager, omSnapshotManager); + } + + Table<String, OmKeyInfo> previousToPrevKeyTable = null; + ReferenceCounted<IOmMetadataReader, SnapshotCache> + rcPrevToPrevOmSnapshot = null; + if (previousToPrevSnapshot != null) { + rcPrevToPrevOmSnapshot = omSnapshotManager.checkForSnapshot( + previousToPrevSnapshot.getVolumeName(), + previousToPrevSnapshot.getBucketName(), + getSnapshotPrefix(previousToPrevSnapshot.getName()), false); + OmSnapshot omPreviousToPrevSnapshot = (OmSnapshot) + rcPrevToPrevOmSnapshot.get(); + + previousToPrevKeyTable = omPreviousToPrevSnapshot + .getMetadataManager() + .getKeyTable(bucketInfo.getBucketLayout()); + } + + String dbBucketKeyForDir = getOzonePathKeyForFso(metadataManager, + currSnapInfo.getVolumeName(), currSnapInfo.getBucketName()); + try (ReferenceCounted<IOmMetadataReader, SnapshotCache> + rcCurrOmSnapshot = omSnapshotManager.checkForSnapshot( + currSnapInfo.getVolumeName(), + currSnapInfo.getBucketName(), + getSnapshotPrefix(currSnapInfo.getName()), + false)) { + + OmSnapshot currOmSnapshot = (OmSnapshot) rcCurrOmSnapshot.get(); + Table<String, OmKeyInfo> snapDeletedDirTable = + currOmSnapshot.getMetadataManager().getDeletedDirTable(); + + try (TableIterator<String, ? extends Table.KeyValue<String, + OmKeyInfo>> deletedDirIterator = snapDeletedDirTable + .iterator(dbBucketKeyForDir)) { + + while (deletedDirIterator.hasNext()) { + Table.KeyValue<String, OmKeyInfo> deletedDirInfo = + deletedDirIterator.next(); + + // For each deleted directory we do an in-memory DFS and + // do a deep clean and exclusive size calculation. + iterateDirectoryTree(deletedDirInfo, volumeId, bucketInfo, + previousSnapshot, previousToPrevSnapshot, + currOmSnapshot, previousKeyTable, prevRenamedTable, + previousToPrevKeyTable, dbBucketKeyForDir); + } + updateDeepCleanSnapshotDir(currSnapInfo.getTableKey()); + if (previousSnapshot != null) { + updateExclusiveSize(previousSnapshot.getTableKey()); + } + } + } + } + } catch (IOException ex) { + LOG.error("Error while running directory deep clean on snapshots." + + " Will retry at next run.", ex); + } + return BackgroundTaskResult.EmptyTaskResult.newResult(); + } + } + + @SuppressWarnings("checkstyle:ParameterNumber") + private void iterateDirectoryTree( + Table.KeyValue<String, OmKeyInfo> deletedDirInfo, long volumeId, + OmBucketInfo bucketInfo, + SnapshotInfo previousSnapshot, + SnapshotInfo previousToPrevSnapshot, + OmSnapshot currOmSnapshot, + Table<String, OmKeyInfo> previousKeyTable, + Table<String, String> prevRenamedTable, + Table<String, OmKeyInfo> previousToPrevKeyTable, + String dbBucketKeyForDir) throws IOException { + + Table<String, OmDirectoryInfo> snapDirTable = + currOmSnapshot.getMetadataManager().getDirectoryTable(); + Table<String, String> snapRenamedTable = + currOmSnapshot.getMetadataManager().getSnapshotRenamedTable(); + Stack<StackNode> stackNodes = + new Stack<>(); + OmDirectoryInfo omDeletedDirectoryInfo = + getDirectoryInfo(deletedDirInfo.getValue()); + String dirPathDbKey = currOmSnapshot.getMetadataManager() + .getOzonePathKey(volumeId, bucketInfo.getObjectID(), + omDeletedDirectoryInfo); + // Stack Init + StackNode topLevelDir = new StackNode(); + topLevelDir.setDirKey(dirPathDbKey); + topLevelDir.setDirValue(omDeletedDirectoryInfo); + stackNodes.add(topLevelDir); + + try ( + TableIterator<String, ? extends Table.KeyValue<String, OmDirectoryInfo>> + directoryIterator = snapDirTable.iterator(dbBucketKeyForDir)) { + + while (!stackNodes.isEmpty()) { + StackNode stackTop = stackNodes.pop(); + String seekDirInDB; + // First process all the files in the current directory + // and then do a DFS for directory. + if (StringUtils.isEmpty(stackTop.getSubDirSeek())) { + processFilesUnderDir(previousSnapshot, + previousToPrevSnapshot, + volumeId, + bucketInfo, + stackTop.getDirValue(), + currOmSnapshot.getMetadataManager(), + snapRenamedTable, + previousKeyTable, + prevRenamedTable, + previousToPrevKeyTable); + seekDirInDB = currOmSnapshot.getMetadataManager() + .getOzonePathKey(volumeId, bucketInfo.getObjectID(), + stackTop.getDirValue().getObjectID(), ""); + directoryIterator.seek(seekDirInDB); + } else { + // When a leaf node is processed, we need to come back in + // the call stack and process the next directories. + seekDirInDB = stackTop.getSubDirSeek(); + directoryIterator.seek(seekDirInDB); + // We need skip to the next sub-directory because we already + // processed the current sub-directory in the previous run. + if (directoryIterator.hasNext()) { + directoryIterator.next(); + } else { + continue; + } + } + + if (directoryIterator.hasNext()) { Review Comment: We need the current directory reference to add the next sub-dir to the stack. So it is simpler to do it in the same iteration than the next iteration. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
