HADOOP-13704. Optimized S3A getContentSummary()

Optimize the scan for s3 by performing a deep tree listing,
inferring directory counts from the paths returned.

Contributed by Ahmar Suhail.

Change-Id: I26ffa8c6f65fd11c68a88d6e2243b0eac6ffd024

hadoop-common-project/hadoop-common/src/site/markdown/filesystem/filesystem.md

@@ -453,6 +453,26 @@ The function `getLocatedFileStatus(FS, d)` is as defined in
 The atomicity and consistency constraints are as for
 `listStatus(Path, PathFilter)`.
 
+
+### `ContentSummary getContentSummary(Path path)`
+
+Given a path return its content summary.
+
+`getContentSummary()` first checks if the given path is a file and if yes, it returns 0 for directory count
+and 1 for file count.
+
+#### Preconditions
+
+    exists(FS, path) else raise FileNotFoundException
+
+#### Postconditions
+
+Returns a `ContentSummary` object with information such as directory count
+and file count for a given path.
+
+The atomicity and consistency constraints are as for
+`listStatus(Path, PathFilter)`.
+
 ### `BlockLocation[] getFileBlockLocations(FileStatus f, int s, int l)`
 
 #### Preconditions

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/fs/contract/AbstractContractContentSummaryTest.java

@@ -0,0 +1,65 @@
+/*
+ * 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.fs.contract;
+
+import org.apache.hadoop.fs.ContentSummary;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+
+import org.assertj.core.api.Assertions;
+import org.junit.Test;
+
+import java.io.FileNotFoundException;
+
+import static org.apache.hadoop.fs.contract.ContractTestUtils.touch;
+import static org.apache.hadoop.test.LambdaTestUtils.intercept;
+
+public abstract class AbstractContractContentSummaryTest extends AbstractFSContractTestBase {
+
+  @Test
+  public void testGetContentSummary() throws Throwable {
+    FileSystem fs = getFileSystem();
+
+    Path parent = path("parent");
+    Path nested = path(parent + "/a/b/c");
+    Path filePath = path(nested + "file.txt");
+
+    fs.mkdirs(parent);
+    fs.mkdirs(nested);
+    touch(getFileSystem(), filePath);
+
+    ContentSummary summary = fs.getContentSummary(parent);
+
+    Assertions.assertThat(summary.getDirectoryCount()).as("Summary " + summary).isEqualTo(4);
+
+    Assertions.assertThat(summary.getFileCount()).as("Summary " + summary).isEqualTo(1);
+  }
+
+  @Test
+  public void testGetContentSummaryIncorrectPath() throws Throwable {
+    FileSystem fs = getFileSystem();
+
+    Path parent = path("parent");
+    Path nested = path(parent + "/a");
+
+    fs.mkdirs(parent);
+
+    intercept(FileNotFoundException.class, () -> fs.getContentSummary(nested));
+  }
+}

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/fs/contract/localfs/TestLocalFSContractContentSummary.java

@@ -0,0 +1,31 @@
+/*
+ * 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.fs.contract.localfs;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.contract.AbstractContractContentSummaryTest;
+import org.apache.hadoop.fs.contract.AbstractFSContract;
+
+public class TestLocalFSContractContentSummary extends AbstractContractContentSummaryTest {
+
+  @Override
+  protected AbstractFSContract createContract(Configuration conf) {
+    return new LocalFSContract(conf);
+  }
+}

hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/S3AFileSystem.java

@@ -3266,9 +3266,9 @@ public class S3AFileSystem extends FileSystem implements StreamCapabilities,
     }
 
     @Override
-    public RemoteIterator<S3AFileStatus> listStatusIterator(final Path path)
-        throws IOException {
-      return S3AFileSystem.this.innerListStatus(path);
+    public RemoteIterator<S3ALocatedFileStatus> listFilesIterator(final Path path,
+        final boolean recursive) throws IOException {
+      return S3AFileSystem.this.innerListFiles(path, recursive, Listing.ACCEPT_ALL_BUT_S3N, null);
     }
   }
 

hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/S3ObjectAttributes.java

@@ -66,7 +66,7 @@ public class S3ObjectAttributes {
   /**
    * Construct from the result of a copy and those parameters
    * which aren't included in an AWS SDK response.
-   * @param path
+   * @param path path
    * @param copyResult copy result.
    * @param serverSideEncryptionAlgorithm current encryption algorithm
    * @param serverSideEncryptionKey any server side encryption key?

hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/impl/GetContentSummaryOperation.java

@@ -20,6 +20,7 @@ package org.apache.hadoop.fs.s3a.impl;
 
 import java.io.FileNotFoundException;
 import java.io.IOException;
+import java.util.HashSet;
 import java.util.Set;
 
 import org.slf4j.Logger;
@@ -34,22 +35,15 @@ import org.apache.hadoop.fs.s3a.S3AFileStatus;
 import org.apache.hadoop.fs.statistics.IOStatistics;
 import org.apache.hadoop.fs.statistics.IOStatisticsSnapshot;
 import org.apache.hadoop.fs.statistics.IOStatisticsSource;
+import org.apache.hadoop.fs.s3a.S3ALocatedFileStatus;
 
 import static org.apache.hadoop.fs.statistics.IOStatisticsSupport.retrieveIOStatistics;
 
 /**
  * GetContentSummary operation.
- * This is based on {@code FileSystem.get#getContentSummary};
- * its still doing sequential treewalk with the efficiency
- * issues.
  *
- * Changes:
- * 1. On the recursive calls there
- * is no probe to see if the path is a file: we know the
- * recursion only happens with a dir.
- * 2. If a subdirectory is not found during the walk, that
- * does not trigger an error. The directory is clearly
- * not part of the content any more.
+ * It is optimized for s3 and performs a deep tree listing,
+ * inferring directory counts from the paths returned.
  *
  * The Operation serves up IOStatistics; this counts
  * the cost of all the list operations, but not the
@@ -122,9 +116,7 @@ public class GetContentSummaryOperation extends
 
   /**
    * Return the {@link ContentSummary} of a given directory.
-   * This is a recursive operation (as the original is);
-   * it'd be more efficient of stack and heap if it managed its
-   * own stack.
+   *
    * @param dir dir to scan
    * @throws FileNotFoundException if the path does not resolve
    * @throws IOException IO failure
@@ -133,34 +125,65 @@ public class GetContentSummaryOperation extends
    * @throws IOException failure
    */
   public ContentSummary getDirSummary(Path dir) throws IOException {
+
     long totalLength = 0;
     long fileCount = 0;
     long dirCount = 1;
-    final RemoteIterator<S3AFileStatus> it
-        = callbacks.listStatusIterator(dir);
+
+    RemoteIterator<S3ALocatedFileStatus> it = callbacks.listFilesIterator(dir, true);
+
+    Set<Path> dirSet = new HashSet<>();
+    Set<Path> pathsTraversed = new HashSet<>();
 
     while (it.hasNext()) {
-      final S3AFileStatus s = it.next();
-      if (s.isDirectory()) {
-        try {
-          ContentSummary c = getDirSummary(s.getPath());
-          totalLength += c.getLength();
-          fileCount += c.getFileCount();
-          dirCount += c.getDirectoryCount();
-        } catch (FileNotFoundException ignored) {
-          // path was deleted during the scan; exclude from
-          // summary.
-        }
-      } else {
-        totalLength += s.getLen();
+      S3ALocatedFileStatus fileStatus = it.next();
+      Path filePath = fileStatus.getPath();
+
+      if (fileStatus.isDirectory() && !filePath.equals(dir)) {
+        dirSet.add(filePath);
+        buildDirectorySet(dirSet, pathsTraversed, dir, filePath.getParent());
+      } else if (!fileStatus.isDirectory()) {
         fileCount += 1;
+        totalLength += fileStatus.getLen();
+        buildDirectorySet(dirSet, pathsTraversed, dir, filePath.getParent());
       }
+
     }
+
     // Add the list's IOStatistics
     iostatistics.aggregate(retrieveIOStatistics(it));
+
     return new ContentSummary.Builder().length(totalLength).
-        fileCount(fileCount).directoryCount(dirCount).
-        spaceConsumed(totalLength).build();
+            fileCount(fileCount).directoryCount(dirCount + dirSet.size()).
+            spaceConsumed(totalLength).build();
+  }
+
+  /***
+   * This method builds the set of all directories found under the base path. We need to do this
+   * because if the directory structure /a/b/c was created with a single mkdirs() call, it is
+   * stored as 1 object in S3 and the list files iterator will only return a single entry /a/b/c.
+   *
+   * We keep track of paths traversed so far to prevent duplication of work. For eg, if we had
+   * a/b/c/file-1.txt and /a/b/c/file-2.txt, we will only recurse over the complete path once
+   * and won't have to do anything for file-2.txt.
+   *
+   * @param dirSet Set of all directories found in the path
+   * @param pathsTraversed Set of all paths traversed so far
+   * @param basePath Path of directory to scan
+   * @param parentPath Parent path of the current file/directory in the iterator
+   */
+  private void buildDirectorySet(Set<Path> dirSet, Set<Path> pathsTraversed, Path basePath,
+      Path parentPath) {
+
+    if (parentPath == null || pathsTraversed.contains(parentPath) || parentPath.equals(basePath)) {
+      return;
+    }
+
+    dirSet.add(parentPath);
+
+    buildDirectorySet(dirSet, pathsTraversed, basePath, parentPath.getParent());
+
+    pathsTraversed.add(parentPath);
   }
 
   /**
@@ -186,23 +209,24 @@ public class GetContentSummaryOperation extends
 
     /**
      * Get the status of a path.
-     * @param path path to probe.
+     *
+     * @param path   path to probe.
      * @param probes probes to exec
      * @return the status
      * @throws IOException failure
      */
     @Retries.RetryTranslated
-    S3AFileStatus probePathStatus(Path path,
-        Set<StatusProbeEnum> probes) throws IOException;
-
-    /**
-     * Incremental list of all entries in a directory.
-     * @param path path of dir
-     * @return an iterator
+    S3AFileStatus probePathStatus(Path path, Set<StatusProbeEnum> probes) throws IOException;
+
+    /***
+     * List all entries under a path.
+     *
+     * @param path
+     * @param recursive if the subdirectories need to be traversed recursively
+     * @return an iterator over the listing.
      * @throws IOException failure
      */
-    RemoteIterator<S3AFileStatus> listStatusIterator(Path path)
+    RemoteIterator<S3ALocatedFileStatus> listFilesIterator(Path path, boolean recursive)
         throws IOException;
-
   }
 }

hadoop-tools/hadoop-aws/src/test/java/org/apache/hadoop/fs/contract/s3a/ITestS3AContractContentSummary.java

@@ -0,0 +1,70 @@
+/*
+ * 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.fs.contract.s3a;
+
+import org.assertj.core.api.Assertions;
+import org.junit.Test;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.ContentSummary;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.fs.contract.AbstractContractContentSummaryTest;
+import org.apache.hadoop.fs.contract.AbstractFSContract;
+import org.apache.hadoop.fs.s3a.S3AFileSystem;
+
+import static org.apache.hadoop.fs.contract.ContractTestUtils.touch;
+
+public class ITestS3AContractContentSummary extends AbstractContractContentSummaryTest {
+
+  @Test
+  public void testGetContentSummaryDir() throws Throwable {
+    describe("getContentSummary on test dir with children");
+    S3AFileSystem fs = getFileSystem();
+    Path baseDir = methodPath();
+
+    // Nested folders created separately will return as separate objects in listFiles()
+    fs.mkdirs(new Path(baseDir, "a"));
+    fs.mkdirs(new Path(baseDir, "a/b"));
+    fs.mkdirs(new Path(baseDir, "a/b/a"));
+
+    // Will return as one object
+    fs.mkdirs(new Path(baseDir, "d/e/f"));
+
+    Path filePath = new Path(baseDir, "a/b/file");
+    touch(fs, filePath);
+
+    // look at path to see if it is a file
+    // it is not: so LIST
+    final ContentSummary summary = fs.getContentSummary(baseDir);
+
+    Assertions.assertThat(summary.getDirectoryCount()).as("Summary " + summary).isEqualTo(7);
+    Assertions.assertThat(summary.getFileCount()).as("Summary " + summary).isEqualTo(1);
+  }
+
+  @Override
+  protected AbstractFSContract createContract(Configuration conf) {
+    return new S3AContract(conf);
+  }
+
+  @Override
+  public S3AFileSystem getFileSystem() {
+    return (S3AFileSystem) super.getFileSystem();
+  }
+
+}

hadoop-tools/hadoop-aws/src/test/java/org/apache/hadoop/fs/s3a/performance/ITestS3AMiscOperationCost.java

@@ -144,8 +144,8 @@ public class ITestS3AMiscOperationCost extends AbstractS3ACostTest {
         with(INVOCATION_GET_CONTENT_SUMMARY, 1),
         withAuditCount(1),
         always(FILE_STATUS_FILE_PROBE    // look at path to see if it is a file
-            .plus(LIST_OPERATION)         // it is not: so LIST
-            .plus(LIST_OPERATION)));       // and a LIST on the child dir
+            .plus(LIST_OPERATION)));         // it is not: so LIST
+
     Assertions.assertThat(summary.getDirectoryCount())
         .as("Summary " + summary)
         .isEqualTo(2);

hadoop-tools/hadoop-aws/src/test/java/org/apache/hadoop/fs/s3a/scale/ITestS3ADirectoryPerformance.java

@@ -19,6 +19,7 @@
 package org.apache.hadoop.fs.s3a.scale;
 
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.ContentSummary;
 import org.apache.hadoop.fs.FileStatus;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.LocatedFileStatus;
@@ -154,6 +155,38 @@ public class ITestS3ADirectoryPerformance extends S3AScaleTestBase {
           listStatusCalls,
           getFileStatusCalls);
 
+      describe("Get content summary for directory");
+
+      NanoTimer getContentSummaryTimer = new NanoTimer();
+
+      ContentSummary rootPathSummary = fs.getContentSummary(scaleTestDir);
+      ContentSummary testPathSummary = fs.getContentSummary(listDir);
+
+      getContentSummaryTimer.end("getContentSummary of %s", created);
+
+      // only two list operations should have taken place
+      print(LOG,
+          metadataRequests,
+          listRequests,
+          listContinueRequests,
+          listStatusCalls,
+          getFileStatusCalls);
+      assertEquals(listRequests.toString(), 2, listRequests.diff());
+      reset(metadataRequests,
+          listRequests,
+          listContinueRequests,
+          listStatusCalls,
+          getFileStatusCalls);
+
+      assertTrue("Root directory count should be > test path",
+          rootPathSummary.getDirectoryCount() > testPathSummary.getDirectoryCount());
+      assertTrue("Root file count should be >= to test path",
+          rootPathSummary.getFileCount() >= testPathSummary.getFileCount());
+      assertEquals("Incorrect directory count", created.getDirCount() + 1,
+          testPathSummary.getDirectoryCount());
+      assertEquals("Incorrect file count", created.getFileCount(),
+          testPathSummary.getFileCount());
+
     } finally {
       describe("deletion");
       // deletion at the end of the run