wchevreuil commented on a change in pull request #3488: URL: https://github.com/apache/hbase/pull/3488#discussion_r675107534
########## File path: hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/DirectStoreFSWriteStrategy.java ########## @@ -0,0 +1,258 @@ +/** + * 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.hbase.regionserver; + +import org.apache.hadoop.fs.FileSystem; +import org.apache.hadoop.fs.Path; +import org.apache.hadoop.hbase.client.ConnectionFactory; +import org.apache.hadoop.hbase.client.RegionInfo; +import org.apache.yetus.audience.InterfaceAudience; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.concurrent.ConcurrentHashMap; + +/** + * <code>HRegionFileSystemWriteStrategy</code> implementation to be used in combination with + * <code>PersistedStoreEngine</code> to avoid renames when splitting and merging regions. + * + * To use it, define the following properties under master configuration: + * 1) <property> + * <name>hbase.hregion.file.write.strategy</name> + * <value>org.apache.hadoop.hbase.regionserver.DirectStoreFSWriteStrategy</value> + * </property> + * 2) <property> + * <name>hbase.hregion.merge.strategy</name> + * <value>org.apache.hadoop.hbase.master.assignment.DirectStoreMergeRegionsStrategy</value> + * </property> + * + * This will create the resulting merging and splitting regions directory straight under + * the table dir, instead of creating it under the temporary ".tmp" or ".merges" dirs, + * as done by the default implementation. + */ [email protected] +public class DirectStoreFSWriteStrategy extends HRegionFileSystemWriteStrategy { + private StoreFilePathAccessor accessor; + private Map<String, Map<String,List<Path>>> regionSplitReferences = new ConcurrentHashMap<>(); + private Map<String, List<Path>> mergeReferences = new HashMap(); + + public DirectStoreFSWriteStrategy(HRegionFileSystem fileSystem) throws IOException { + super(fileSystem); + this.accessor = StoreFileTrackingUtils.createStoreFilePathAccessor(fileSystem.conf, + ConnectionFactory.createConnection(fileSystem.conf)); + } + + /** + * The parent directory where to create the splits dirs is + * the table directory itself, in this case. + * @return Path representing the table directory. + */ + @Override + public Path getParentSplitsDir() { + return fileSystem.getTableDir(); + } + + /** + * The parent directory where to create the merge dir is + * the table directory itself, in this case. + * @return Path representing the table directory. + */ + @Override + public Path getParentMergesDir() { + return fileSystem.getTableDir(); + } + + /** + * Creates the directories for the respective split daughters directly under the + * table directory, instead of default behaviour of doing it under temp dirs, initially. + * @param daughterA the first half of the split region + * @param daughterB the second half of the split region + * + * @throws IOException if directories creation fails. + */ + @Override + public void createSplitsDir(RegionInfo daughterA, RegionInfo daughterB) + throws IOException { + Path splitdir = getParentSplitsDir(); + // splitDir doesn't exists now. No need to do an exists() call for it. + if (!fileSystem.getFileSystem().exists(splitdir)) { + throw new IOException("Table dir for splitting region not found: " + splitdir); + } + Path daughterADir = getSplitsDir(daughterA); + if (!fileSystem.createDir(daughterADir)) { + throw new IOException("Failed create of " + daughterADir); + } + Path daughterBDir = getSplitsDir(daughterB); + if (!fileSystem.createDir(daughterBDir)) { + throw new IOException("Failed create of " + daughterBDir); + } + } + + /** + * Just validates that merges parent, the actual table dir in this case, exists. + * @throws IOException if table dir doesn't exist. + */ + @Override + public void createMergesDir() throws IOException { + //When writing directly, avoiding renames, merges parent is the table dir itself, so it + // should exist already, so just validate it exist then do nothing + Path mergesdir = getParentMergesDir(); + if (!fileSystem.fs.exists(mergesdir)) { + throw new IOException("Table dir for merging region not found: " + mergesdir); + } Review comment: I know, this is very confusing. In the original implementation, `MergeTableRegionsProcedure.createMergedRegion` creates an `HRegionFileSystem` instance for the first region being merged, then calls `HRegionFileSystem.createMergesDir`, which creates a `.merges` temp dir under this region dir. So what has always been created by `HRegionFileSystem.createMergesDir` was not the new region dir itself, but the temporary `.merges` dir to store the merged files during the merging process. When dealing with the `DirectStoreMergeRegionsStrategy`, there's really nothing to create here, as the `HRegionFileSystem` is already representing the new resulting merging region. -- 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]
