Recursive Delete on BlobContainer (#43281) (#43920)

This is a prerequisite of #42189:

* Add directory delete method to blob container specific to each implementation:
  * Some notes on the implementations:
       * AWS + GCS: We can simply exploit the fact that both AWS and GCS return blobs lexicographically ordered which allows us to simply delete in the same order that we receive the blobs from the listing request. For AWS this simply required listing without the delimiter setting (so we get a deep listing) and for GCS the same behavior is achieved by not using the directory mode on the listing invocation. The nice thing about this is, that even for very large numbers of blobs the memory requirements are now capped nicely since we go page by page when deleting.
       * For Azure I extended the parallelization to the listing calls as well and made it work recursively. I verified that this works with thread count `1` since we only block once in the initial thread and then fan out to a "graph" of child listeners that never block.
       * HDFS and FS are trivial since we have directory delete methods available for them
* Enhances third party tests to ensure the new functionality works (I manually ran them for all cloud providers)

Author: original-brownbear

modules/repository-url/src/main/java/org/elasticsearch/common/blobstore/url/URLBlobContainer.java

@@ -96,6 +96,11 @@ public void deleteBlob(String blobName) throws IOException {
         throw new UnsupportedOperationException("URL repository is read only");
     }
 
+    @Override
+    public void delete() {
+        throw new UnsupportedOperationException("URL repository is read only");
+    }
+
     /**
      * This operation is not supported by URLBlobContainer
      */

plugins/repository-azure/src/main/java/org/elasticsearch/repositories/azure/AzureBlobContainer.java

@@ -23,6 +23,7 @@
 import com.microsoft.azure.storage.StorageException;
 import org.apache.logging.log4j.LogManager;
 import org.apache.logging.log4j.Logger;
+import org.elasticsearch.action.ActionListener;
 import org.elasticsearch.action.ActionRunnable;
 import org.elasticsearch.action.support.GroupedActionListener;
 import org.elasticsearch.action.support.PlainActionFuture;
@@ -38,7 +39,6 @@
 import java.net.HttpURLConnection;
 import java.net.URISyntaxException;
 import java.nio.file.NoSuchFileException;
-import java.util.Collection;
 import java.util.List;
 import java.util.Map;
 import java.util.concurrent.ExecutorService;
@@ -126,24 +126,35 @@ public void deleteBlob(String blobName) throws IOException {
         }
     }
 
+    @Override
+    public void delete() throws IOException {
+        try {
+            blobStore.deleteBlobDirectory(keyPath, threadPool.executor(AzureRepositoryPlugin.REPOSITORY_THREAD_POOL_NAME));
+        } catch (URISyntaxException | StorageException e) {
+            throw new IOException(e);
+        }
+    }
+
     @Override
     public void deleteBlobsIgnoringIfNotExists(List<String> blobNames) throws IOException {
+        final PlainActionFuture<Void> result = PlainActionFuture.newFuture();
         if (blobNames.isEmpty()) {
-            return;
-        }
-        final PlainActionFuture<Collection<Void>> result = PlainActionFuture.newFuture();
-        final GroupedActionListener<Void> listener = new GroupedActionListener<>(result, blobNames.size());
-        final ExecutorService executor = threadPool.executor(AzureRepositoryPlugin.REPOSITORY_THREAD_POOL_NAME);
-        // Executing deletes in parallel since Azure SDK 8 is using blocking IO while Azure does not provide a bulk delete API endpoint.
-        // TODO: Upgrade to newer non-blocking Azure SDK 11 and execute delete requests in parallel that way.
-        for (String blobName : blobNames) {
-            executor.submit(new ActionRunnable<Void>(listener) {
-                @Override
-                protected void doRun() throws IOException {
-                    deleteBlobIgnoringIfNotExists(blobName);
-                    listener.onResponse(null);
-                }
-            });
+            result.onResponse(null);
+        } else {
+            final GroupedActionListener<Void> listener =
+                new GroupedActionListener<>(ActionListener.map(result, v -> null), blobNames.size());
+            final ExecutorService executor = threadPool.executor(AzureRepositoryPlugin.REPOSITORY_THREAD_POOL_NAME);
+            // Executing deletes in parallel since Azure SDK 8 is using blocking IO while Azure does not provide a bulk delete API endpoint
+            // TODO: Upgrade to newer non-blocking Azure SDK 11 and execute delete requests in parallel that way.
+            for (String blobName : blobNames) {
+                executor.execute(new ActionRunnable<Void>(listener) {
+                    @Override
+                    protected void doRun() throws IOException {
+                        deleteBlobIgnoringIfNotExists(blobName);
+                        listener.onResponse(null);
+                    }
+                });
+            }
         }
         try {
             result.actionGet();

plugins/repository-azure/src/main/java/org/elasticsearch/repositories/azure/AzureBlobStore.java

@@ -36,6 +36,7 @@
 import java.nio.file.FileAlreadyExistsException;
 import java.util.Collections;
 import java.util.Map;
+import java.util.concurrent.Executor;
 import java.util.function.Function;
 import java.util.stream.Collectors;
 
@@ -91,6 +92,10 @@ public void deleteBlob(String blob) throws URISyntaxException, StorageException
         service.deleteBlob(clientName, container, blob);
     }
 
+    public void deleteBlobDirectory(String path, Executor executor) throws URISyntaxException, StorageException, IOException {
+        service.deleteBlobDirectory(clientName, container, path, executor);
+    }
+
     public InputStream getInputStream(String blob) throws URISyntaxException, StorageException, IOException {
         return service.getInputStream(clientName, container, blob);
     }

plugins/repository-azure/src/main/java/org/elasticsearch/repositories/azure/AzureStorageService.java

@@ -40,6 +40,7 @@
 import org.apache.logging.log4j.LogManager;
 import org.apache.logging.log4j.Logger;
 import org.apache.logging.log4j.message.ParameterizedMessage;
+import org.elasticsearch.action.support.PlainActionFuture;
 import org.elasticsearch.common.blobstore.BlobMetaData;
 import org.elasticsearch.common.blobstore.BlobPath;
 import org.elasticsearch.common.blobstore.support.PlainBlobMetaData;
@@ -49,6 +50,7 @@
 import org.elasticsearch.common.settings.SettingsException;
 import org.elasticsearch.common.unit.ByteSizeUnit;
 import org.elasticsearch.common.unit.ByteSizeValue;
+import org.elasticsearch.common.util.concurrent.AbstractRunnable;
 
 import java.io.IOException;
 import java.io.InputStream;
@@ -57,11 +59,15 @@
 import java.net.URISyntaxException;
 import java.nio.file.FileAlreadyExistsException;
 import java.security.InvalidKeyException;
+import java.util.ArrayList;
+import java.util.Collection;
 import java.util.Collections;
 import java.util.EnumSet;
 import java.util.HashSet;
 import java.util.Map;
 import java.util.Set;
+import java.util.concurrent.Executor;
+import java.util.concurrent.atomic.AtomicLong;
 import java.util.function.Supplier;
 
 import static java.util.Collections.emptyMap;
@@ -187,6 +193,50 @@ public void deleteBlob(String account, String container, String blob) throws URI
         });
     }
 
+    void deleteBlobDirectory(String account, String container, String path, Executor executor)
+            throws URISyntaxException, StorageException, IOException {
+        final Tuple<CloudBlobClient, Supplier<OperationContext>> client = client(account);
+        final CloudBlobContainer blobContainer = client.v1().getContainerReference(container);
+        final Collection<Exception> exceptions = Collections.synchronizedList(new ArrayList<>());
+        final AtomicLong outstanding = new AtomicLong(1L);
+        final PlainActionFuture<Void> result = PlainActionFuture.newFuture();
+        SocketAccess.doPrivilegedVoidException(() -> {
+            for (final ListBlobItem blobItem : blobContainer.listBlobs(path, true)) {
+                // uri.getPath is of the form /container/keyPath.* and we want to strip off the /container/
+                // this requires 1 + container.length() + 1, with each 1 corresponding to one of the /
+                final String blobPath = blobItem.getUri().getPath().substring(1 + container.length() + 1);
+                outstanding.incrementAndGet();
+                executor.execute(new AbstractRunnable() {
+                    @Override
+                    protected void doRun() throws Exception {
+                        deleteBlob(account, container, blobPath);
+                    }
+
+                    @Override
+                    public void onFailure(Exception e) {
+                        exceptions.add(e);
+                    }
+
+                    @Override
+                    public void onAfter() {
+                        if (outstanding.decrementAndGet() == 0) {
+                            result.onResponse(null);
+                        }
+                    }
+                });
+            }
+        });
+        if (outstanding.decrementAndGet() == 0) {
+            result.onResponse(null);
+        }
+        result.actionGet();
+        if (exceptions.isEmpty() == false) {
+            final IOException ex = new IOException("Deleting directory [" + path + "] failed");
+            exceptions.forEach(ex::addSuppressed);
+            throw ex;
+        }
+    }
+
     public InputStream getInputStream(String account, String container, String blob)
         throws URISyntaxException, StorageException, IOException {
         final Tuple<CloudBlobClient, Supplier<OperationContext>> client = client(account);

plugins/repository-gcs/src/main/java/org/elasticsearch/repositories/gcs/GoogleCloudStorageBlobContainer.java

@@ -86,6 +86,11 @@ public void deleteBlob(String blobName) throws IOException {
         blobStore.deleteBlob(buildKey(blobName));
     }
 
+    @Override
+    public void delete() throws IOException {
+        blobStore.deleteDirectory(path().buildAsString());
+    }
+
     @Override
     public void deleteBlobsIgnoringIfNotExists(List<String> blobNames) throws IOException {
         blobStore.deleteBlobsIgnoringIfNotExists(blobNames.stream().map(this::buildKey).collect(Collectors.toList()));

plugins/repository-gcs/src/main/java/org/elasticsearch/repositories/gcs/GoogleCloudStorageBlobStore.java

@@ -19,6 +19,7 @@
 
 package org.elasticsearch.repositories.gcs;
 
+import com.google.api.gax.paging.Page;
 import com.google.cloud.BatchResult;
 import com.google.cloud.ReadChannel;
 import com.google.cloud.WriteChannel;
@@ -306,6 +307,23 @@ void deleteBlob(String blobName) throws IOException {
         }
     }
 
+    /**
+     * Deletes the given path and all its children.
+     *
+     * @param pathStr Name of path to delete
+     */
+    void deleteDirectory(String pathStr) throws IOException {
+        SocketAccess.doPrivilegedVoidIOException(() -> {
+            Page<Blob> page = client().get(bucketName).list(BlobListOption.prefix(pathStr));
+            do {
+                final Collection<String> blobsToDelete = new ArrayList<>();
+                page.getValues().forEach(b -> blobsToDelete.add(b.getName()));
+                deleteBlobsIgnoringIfNotExists(blobsToDelete);
+                page = page.getNextPage();
+            } while (page != null);
+        });
+    }
+
     /**
      * Deletes multiple blobs from the specific bucket using a batch request
      *

plugins/repository-hdfs/src/main/java/org/elasticsearch/repositories/hdfs/HdfsBlobContainer.java

@@ -78,6 +78,11 @@ public void deleteBlob(String blobName) throws IOException {
         }
     }
 
+    @Override
+    public void delete() throws IOException {
+        store.execute(fileContext -> fileContext.delete(path, true));
+    }
+
     @Override
     public InputStream readBlob(String blobName) throws IOException {
         // FSDataInputStream does buffering internally

plugins/repository-s3/src/main/java/org/elasticsearch/repositories/s3/S3BlobContainer.java

@@ -54,6 +54,7 @@
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.HashMap;
+import java.util.HashSet;
 import java.util.List;
 import java.util.Map;
 import java.util.Set;
@@ -130,12 +131,53 @@ public void deleteBlob(String blobName) throws IOException {
         deleteBlobIgnoringIfNotExists(blobName);
     }
 
+    @Override
+    public void delete() throws IOException {
+        try (AmazonS3Reference clientReference = blobStore.clientReference()) {
+            ObjectListing prevListing = null;
+            while (true) {
+                ObjectListing list;
+                if (prevListing != null) {
+                    final ObjectListing finalPrevListing = prevListing;
+                    list = SocketAccess.doPrivileged(() -> clientReference.client().listNextBatchOfObjects(finalPrevListing));
+                } else {
+                    final ListObjectsRequest listObjectsRequest = new ListObjectsRequest();
+                    listObjectsRequest.setBucketName(blobStore.bucket());
+                    listObjectsRequest.setPrefix(keyPath);
+                    list = SocketAccess.doPrivileged(() -> clientReference.client().listObjects(listObjectsRequest));
+                }
+                final List<String> blobsToDelete =
+                    list.getObjectSummaries().stream().map(S3ObjectSummary::getKey).collect(Collectors.toList());
+                if (list.isTruncated()) {
+                    doDeleteBlobs(blobsToDelete, false);
+                    prevListing = list;
+                } else {
+                    final List<String> lastBlobsToDelete = new ArrayList<>(blobsToDelete);
+                    lastBlobsToDelete.add(keyPath);
+                    doDeleteBlobs(lastBlobsToDelete, false);
+                    break;
+                }
+            }
+        } catch (final AmazonClientException e) {
+            throw new IOException("Exception when deleting blob container [" + keyPath + "]", e);
+        }
+    }
+
     @Override
     public void deleteBlobsIgnoringIfNotExists(List<String> blobNames) throws IOException {
+        doDeleteBlobs(blobNames, true);
+    }
+
+    private void doDeleteBlobs(List<String> blobNames, boolean relative) throws IOException {
         if (blobNames.isEmpty()) {
             return;
         }
-        final Set<String> outstanding = blobNames.stream().map(this::buildKey).collect(Collectors.toSet());
+        final Set<String> outstanding;
+        if (relative) {
+            outstanding = blobNames.stream().map(this::buildKey).collect(Collectors.toSet());
+        } else {
+            outstanding = new HashSet<>(blobNames);
+        }
         try (AmazonS3Reference clientReference = blobStore.clientReference()) {
             // S3 API only allows 1k blobs per delete so we split up the given blobs into requests of max. 1k deletes
             final List<DeleteObjectsRequest> deleteRequests = new ArrayList<>();

plugins/repository-s3/src/test/java/org/elasticsearch/repositories/s3/S3RepositoryThirdPartyTests.java

@@ -88,4 +88,11 @@ protected void assertChildren(BlobPath path, Collection<String> children) throws
         // to become consistent.
         assertBusy(() -> super.assertChildren(path, children), 10L, TimeUnit.MINUTES);
     }
+
+    @Override
+    protected void assertDeleted(BlobPath path, String name) throws Exception {
+        // AWS S3 is eventually consistent so we retry for 10 minutes assuming a list operation will never take longer than that
+        // to become consistent.
+        assertBusy(() -> super.assertDeleted(path, name), 10L, TimeUnit.MINUTES);
+    }
 }

server/src/main/java/org/elasticsearch/common/blobstore/BlobContainer.java

@@ -109,6 +109,12 @@ public interface BlobContainer {
      */
     void deleteBlob(String blobName) throws IOException;
 
+    /**
+     * Deletes this container and all its contents from the repository.
+     * @throws IOException on failure
+     */
+    void delete() throws IOException;
+
     /**
      * Deletes the blobs with given names. Unlike {@link #deleteBlob(String)} this method will not throw an exception
      * when one or multiple of the given blobs don't exist and simply ignore this case.

server/src/main/java/org/elasticsearch/common/blobstore/BlobPath.java

@@ -19,6 +19,8 @@
 
 package org.elasticsearch.common.blobstore;
 
+import org.elasticsearch.common.Nullable;
+
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.Iterator;
@@ -68,6 +70,20 @@ public String buildAsString() {
         return p + SEPARATOR;
     }
 
+    /**
+     * Returns this path's parent path.
+     *
+     * @return Parent path or {@code null} if there is none
+     */
+    @Nullable
+    public BlobPath parent() {
+        if (paths.isEmpty()) {
+            return null;
+        } else {
+            return new BlobPath(new ArrayList<>(paths.subList(0, paths.size() - 1)));
+        }
+    }
+
     @Override
     public String toString() {
         StringBuilder sb = new StringBuilder();

server/src/main/java/org/elasticsearch/common/blobstore/fs/FsBlobContainer.java

@@ -122,6 +122,11 @@ public FileVisitResult postVisitDirectory(Path dir, IOException exc) throws IOEx
         }
     }
 
+    @Override
+    public void delete() throws IOException {
+        IOUtils.rm(path);
+    }
+
     @Override
     public boolean blobExists(String blobName) {
         return Files.exists(path.resolve(blobName));

server/src/test/java/org/elasticsearch/snapshots/mockstore/BlobContainerWrapper.java

@@ -64,6 +64,11 @@ public void deleteBlob(String blobName) throws IOException {
         delegate.deleteBlob(blobName);
     }
 
+    @Override
+    public void delete() throws IOException {
+        delegate.delete();
+    }
+
     @Override
     public void deleteBlobIgnoringIfNotExists(final String blobName) throws IOException {
         delegate.deleteBlobIgnoringIfNotExists(blobName);