ilooner commented on a change in pull request #1296: DRILL-5365: Prevent plugin config from changing default fs. Make DrillFileSystem Immutable. URL: https://github.com/apache/drill/pull/1296#discussion_r204148883
########## File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/dfs/DrillFileSystemCache.java ########## @@ -0,0 +1,99 @@ +/* + * 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.drill.exec.store.dfs; + +import com.google.common.base.Preconditions; +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.fs.FileSystem; + +import java.io.IOException; +import java.util.HashMap; +import java.util.Map; + +/** + * <h4>Motivation</h4> + * <p> + * This cache is intended to work around the bugs in the {@link org.apache.hadoop.fs.FileSystem} static cache (DRILL-5365). Specifically, as of Hadoop 2.7.x the + * {@link org.apache.hadoop.fs.FileSystem} cache has the following bad behavior: + * </p> + * <ul> + * <li> + * The {@link org.apache.hadoop.conf.Configuration} object is not considered when constructing keys for the {@link org.apache.hadoop.fs.FileSystem} cache of + * {@link org.apache.hadoop.fs.FileSystem} objects. + * </li> + * <li> + * The {@link org.apache.hadoop.fs.FileSystem} cache does not honor the <b>fs.default.name</b> property when constructing keys, only <b>fs.defaultFS</b> is used to construct + * keys in the cache. + * </li> + * </ul> + * + * <h4>Usage</h4> + * + * <ul> + * <li> + * A prerequisite for usage is that all {@link org.apache.hadoop.conf.Configuration} objects are normalized with + * {@link org.apache.drill.exec.store.dfs.DrillFileSystem#normalize(Configuration)}. + * </li> + * <li> + * This cache should only be used from {@link org.apache.drill.exec.store.dfs.DrillFileSystem}. + * </li> + * </ul> + * + * <h4>TODO</h4> + * + * <ul> + * <li> + * Drill currently keeps a {@link org.apache.drill.exec.store.dfs.DrillFileSystem} open indefinitely. This will be corrected + * in DRILL-6608. As a result this cache currently has no methods to remove {@link org.apache.hadoop.fs.FileSystem} objects + * after they are created. + * </li> + * </ul> + */ +class DrillFileSystemCache { + private Map<Map<String, String>, FileSystem> cache = new HashMap<>(); + + /** + * If a {@link org.apache.hadoop.fs.FileSystem} object corresponding to the given {@link org.apache.hadoop.conf.Configuration} + * exists in the cache, then it is returned. If no corresponding {@link org.apache.hadoop.fs.FileSystem} exist, then it is created, + * added to the cache, and returned. + * @param configuration The {@link org.apache.hadoop.conf.Configuration} corresponding to the desired {@link org.apache.hadoop.fs.FileSystem} + * object. It is expected that this configuration is first normalized with {@link org.apache.drill.exec.store.dfs.DrillFileSystem#normalize(Configuration)}. + * @return The {@link org.apache.hadoop.fs.FileSystem} object corresponding to he given {@link org.apache.hadoop.conf.Configuration}. + * @throws IOException An error when creating the desired {@link org.apache.hadoop.fs.FileSystem} object. + */ + protected synchronized FileSystem get(final Configuration configuration) throws IOException { + final Map<String, String> map = new HashMap<>(configToMap(configuration)); + + if (!cache.containsKey(map)) { + cache.put(map, FileSystem.newInstance(configuration)); + } + + return cache.get(map); + } + + static Map<String, String> configToMap(final Configuration configuration) { + Preconditions.checkNotNull(configuration); + final Map<String, String> map = new HashMap<>(); + + for (Map.Entry<String, String> entry: configuration) { + map.put(entry.getKey().trim(), entry.getValue()); Review comment: - The code for CACHE.get calls the Key(uri, conf) constructor https://github.com/apache/hadoop/blob/branch-2.7.1/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileSystem.java#L2667 . That constructor does not initialize the ugi object, so it is not used as part of the key. https://github.com/apache/hadoop/blob/branch-2.7.1/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileSystem.java#L2801 The UGI object is only used when CACHE.getUnique is called. In that method the Key(uri, conf, unique) constructor is called https://github.com/apache/hadoop/blob/branch-2.7.1/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileSystem.java#L2673 . Only in that case does the Key constructor call UGI.getCurrentUser() . https://github.com/apache/hadoop/blob/branch-2.7.1/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileSystem.java#L2801 . So if not using the UGI object is a security issue, it is a pre-existing issue. Please double check the code I referenced to make sure my analysis is correct. - You are right I had the deprecated and non-deprecated property names mixed up. Thank you very much for catching this! - While there is static conf field in the FileSystem class, no one seems to use it. For example DistrubutedFileSystem calls the setConf method https://github.com/apache/hadoop/blob/branch-2.7.1/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DistributedFileSystem.java#L137, which sets a conf variable only for that instance of the class https://github.com/apache/hadoop/blob/branch-2.7.1/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/conf/Configured.java (DistributedFileSystem extends FileSystem which extends Configured). In fact as far as I can tell from my IDE the static conf variable is never used. Whenever the FileSystem class uses the configuration it calls getConf() which is a method of Configured and which returns the configuration that is stored on a per instance basis. So I think the configuration should still be used in the key for caching. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on 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
