You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by GitBox <gi...@apache.org> on 2022/12/23 17:20:56 UTC
[GitHub] [flink] anoopsjohn commented on a diff in pull request #21508: [FLINK-30128] Introduce Azure Data Lake Gen2 APIs in the Hadoop Recoverable path
anoopsjohn commented on code in PR #21508:
URL: https://github.com/apache/flink/pull/21508#discussion_r1051328889
##########
flink-filesystems/flink-azure-fs-hadoop/src/main/java/org/apache/flink/fs/azurefs/AzureBlobFsRecoverableDataOutputStream.java:
##########
@@ -0,0 +1,299 @@
+/*
+ * 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.flink.fs.azurefs;
+
+import org.apache.flink.annotation.Internal;
+import org.apache.flink.core.fs.RecoverableFsDataOutputStream;
+import org.apache.flink.core.fs.RecoverableWriter;
+import org.apache.flink.runtime.fs.hdfs.HadoopFsRecoverable;
+
+import org.apache.hadoop.fs.FSDataInputStream;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.FileNotFoundException;
+import java.io.IOException;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+/**
+ * An implementation of the {@link RecoverableFsDataOutputStream} for AzureBlob's file system
+ * abstraction.
+ */
+@Internal
+public class AzureBlobFsRecoverableDataOutputStream extends RecoverableFsDataOutputStream {
+
+ private static final Logger LOG =
+ LoggerFactory.getLogger(AzureBlobFsRecoverableDataOutputStream.class);
+
+ private final FileSystem fs;
+
+ private final Path targetFile;
+
+ private final Path tempFile;
+
+ private final FSDataOutputStream out;
+
+ // Not final to override in tests
+ public static int minBufferLength = 2097152;
+
+ // init to 0. When ever recovery is done add this to the pos.
+ private long initialFileSize = 0;
+
+ public AzureBlobFsRecoverableDataOutputStream(FileSystem fs, Path targetFile, Path tempFile)
+ throws IOException {
+ this.fs = checkNotNull(fs);
+ this.targetFile = checkNotNull(targetFile);
+ LOG.debug("The targetFile is {}", targetFile.getName());
+ this.tempFile = checkNotNull(tempFile);
+ LOG.debug("The tempFile is {}", tempFile.getName());
+ this.out = fs.create(tempFile);
+ }
+
+ public AzureBlobFsRecoverableDataOutputStream(FileSystem fs, HadoopFsRecoverable recoverable)
+ throws IOException {
+ this.fs = checkNotNull(fs);
+ this.targetFile = checkNotNull(recoverable.targetFile());
+ this.tempFile = checkNotNull(recoverable.tempFile());
+ long len = fs.getFileStatus(tempFile).getLen();
+ LOG.info("The recoverable offset is {} and the file len is {}", recoverable.offset(), len);
+ // Happens when we recover from a previously committed offset. Otherwise this is not
+ // really needed
+ if (len > recoverable.offset()) {
+ truncate(fs, recoverable);
+ } else if (len < recoverable.offset()) {
+ LOG.error(
+ "Temp file length {} is less than the expected recoverable offset {}",
+ len,
+ recoverable.offset());
+ throw new IOException(
+ "Unable to create recoverable outputstream as length of file "
+ + len
+ + " is less than "
+ + "recoverable offset "
+ + recoverable.offset());
+ }
+ // In ABFS when we try to append we don't account for the initial file size like we do in
+ // DFS.
+ // So we explicitly store this and when we do a persist call we make use of it.
+ initialFileSize = fs.getFileStatus(tempFile).getLen();
+ out = fs.append(tempFile);
+ LOG.debug("Created a new OS for appending {}", tempFile);
+ }
+
+ private void truncate(FileSystem fs, HadoopFsRecoverable recoverable) throws IOException {
+ Path renameTempPath = new Path(tempFile.toString() + ".rename");
+ try {
+ LOG.info(
+ "Creating the temp rename file {} for truncating the tempFile {}",
+ renameTempPath,
+ tempFile);
+ FSDataOutputStream fsDataOutputStream = fs.create(renameTempPath);
+ LOG.info("Opening the tempFile {} for truncate", tempFile);
+ FSDataInputStream fsDis = fs.open(tempFile);
+ // 2 MB buffers. TODO : Make this configurable
+ long remaining = recoverable.offset();
+ byte[] buf = null;
+ long dataWritten = 0;
+ while (remaining != 0) {
+ if (minBufferLength < remaining) {
+ buf = new byte[minBufferLength];
+ } else {
+ buf = new byte[(int) remaining];
+ }
+ fsDis.read(buf, 0, buf.length);
+ remaining -= buf.length;
+ LOG.info("Bytes remaining to read {}", remaining);
+ fsDataOutputStream.write(buf);
+ dataWritten += buf.length;
+ LOG.info("Successfully wrote {} bytes of data {}", dataWritten);
+ }
+ // TODO : Support intermediate flush?
+ LOG.info("Closing the temp rename file {}", renameTempPath);
+ fsDataOutputStream.close();
+ } catch (IOException e) {
+ LOG.error(
+ "Unable to recover. Exception while trying to truncate the temp file {}",
+ tempFile);
+ // We cannot recover. This we can control if user does not want this??
+ throw e;
+ }
+ try {
+ LOG.info("Deleting the actual temp file {}", tempFile);
+ fs.delete(tempFile, false);
+ } catch (IOException e) {
+ LOG.error("Unable to recover. Error while deleting the temp file {}", tempFile);
+ // unable to recover.
+ throw e;
+ }
+ rename(fs, renameTempPath);
+ }
+
+ private void rename(FileSystem fs, Path renameTempPath) throws IOException {
+ LOG.info("Renaming the temp rename file {} back to tempFile {}", renameTempPath, tempFile);
+ try {
+ boolean result = fs.rename(renameTempPath, tempFile);
+ if (!result) {
+ LOG.error(
+ "Unable to recover. Rename operation failed {} to {}",
+ renameTempPath,
+ tempFile);
+ throw new IOException("Unable to recover. Rename operation failed");
+ } else {
+ LOG.info("Rename was successful");
+ }
+ } catch (IOException e) {
+ LOG.error(
+ "Unable to recover. Renaming of tempFile did not happen after truncating {} to {}",
+ renameTempPath,
+ tempFile);
+ throw e;
+ }
+ }
+
+ @Override
+ public void write(byte[] b, int off, int len) throws IOException {
+ out.write(b, off, len);
+ }
+
+ @Override
+ public long getPos() throws IOException {
+ return out.getPos();
+ }
+
+ @Override
+ public void write(int b) throws IOException {
+ out.write(b);
+ }
+
+ @Override
+ public void flush() throws IOException {
+ out.hflush();
+ }
+
+ @Override
+ public void sync() throws IOException {
+ out.hflush();
Review Comment:
Minor: Here we handle ABFS OS and you can call hsync alone. Anyways both APIs doing same in ABFS case.
##########
flink-filesystems/flink-azure-fs-hadoop/src/main/java/org/apache/flink/fs/azurefs/AzureBlobFsRecoverableDataOutputStream.java:
##########
@@ -0,0 +1,312 @@
+/*
+ * 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.flink.fs.azurefs;
+
+import org.apache.flink.annotation.Internal;
+import org.apache.flink.core.fs.RecoverableFsDataOutputStream;
+import org.apache.flink.core.fs.RecoverableWriter;
+import org.apache.flink.runtime.fs.hdfs.HadoopFsRecoverable;
+
+import org.apache.hadoop.fs.FSDataInputStream;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.FileNotFoundException;
+import java.io.IOException;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+/**
+ * An implementation of the {@link RecoverableFsDataOutputStream} for AzureBlob's file system
+ * abstraction.
+ */
+@Internal
+public class AzureBlobFsRecoverableDataOutputStream extends RecoverableFsDataOutputStream {
+
+ private static final Logger LOG =
+ LoggerFactory.getLogger(AzureBlobFsRecoverableDataOutputStream.class);
+
+ private final FileSystem fs;
+
+ private final Path targetFile;
+
+ private final Path tempFile;
+
+ private final FSDataOutputStream out;
+
+ // Not final to override in tests
+ public static int minBufferLength = 2097152;
+
+ // init to 0. When ever recovery is done add this to the pos.
+ private long initialFileSize = 0;
+
+ AzureBlobFsRecoverableDataOutputStream(FileSystem fs, Path targetFile, Path tempFile)
+ throws IOException {
+ this.fs = checkNotNull(fs);
+ this.targetFile = checkNotNull(targetFile);
+ LOG.debug("The targetFile is {}", targetFile.getName());
+ this.tempFile = checkNotNull(tempFile);
+ LOG.debug("The tempFile is {}", tempFile.getName());
+ this.out = fs.create(tempFile);
+ }
+
+ AzureBlobFsRecoverableDataOutputStream(FileSystem fs, HadoopFsRecoverable recoverable)
+ throws IOException {
+ this.fs = checkNotNull(fs);
+ this.targetFile = checkNotNull(recoverable.targetFile());
+ this.tempFile = checkNotNull(recoverable.tempFile());
+ long len = fs.getFileStatus(tempFile).getLen();
+ LOG.info("The recoverable offset is {} and the file len is {}", recoverable.offset(), len);
+ // Happens when we recover from a previously committed offset. Otherwise this is not
+ // really needed
+ if (len > recoverable.offset()) {
+ truncate(fs, recoverable);
+ } else if (len < recoverable.offset()) {
+ LOG.error(
+ "Temp file length {} is less than the expected recoverable offset {}",
+ len,
+ recoverable.offset());
+ throw new IOException(
+ "Unable to create recoverable outputstream as length of file "
+ + len
+ + " is less than "
+ + "recoverable offset "
+ + recoverable.offset());
+ }
+ out = fs.append(tempFile);
+ if (out.getPos() == 0) {
+ // In ABFS when we try to append we don't account for the initial file size like we do
+ // in DFS.
+ // So we explicitly store this and when we do a persist call we make use of it.
+ // This we have raised a bug in ABFS hadoop driver side. Once fixed this will not be
+ // needed. So it should be ok to put this in side the 'if' check.
+ initialFileSize = fs.getFileStatus(tempFile).getLen();
+ }
+ LOG.debug("Created a new OS for appending {}", tempFile);
+ }
+
+ private void truncate(FileSystem fs, HadoopFsRecoverable recoverable) throws IOException {
+ Path renameTempPath = new Path(tempFile.toString() + ".rename");
+ try {
+ LOG.info(
+ "Creating the temp rename file {} for truncating the tempFile {}",
+ renameTempPath,
+ tempFile);
+ FSDataOutputStream fsDataOutputStream = fs.create(renameTempPath);
+ LOG.info("Opening the tempFile {} for truncate", tempFile);
+ FSDataInputStream fsDis = fs.open(tempFile);
+ // 2 MB buffers. TODO : Make this configurable
+ long remaining = recoverable.offset();
+ byte[] buf = null;
+ long dataWritten = 0;
+ int readBytes = -1;
+ while (remaining != 0) {
+ if (minBufferLength < remaining) {
+ buf = new byte[minBufferLength];
+ } else {
+ buf = new byte[(int) remaining];
+ }
+ readBytes = fsDis.read(buf, 0, buf.length);
+ if (readBytes != -1) {
+ remaining -= readBytes;
+ LOG.info("Bytes remaining to read {}", remaining);
+ fsDataOutputStream.write(buf, 0, readBytes);
+ dataWritten += readBytes;
+ LOG.info("Successfully wrote {} bytes of data", dataWritten);
+ } else {
+ LOG.debug("Reached the end of the file");
+ remaining = 0;
+ }
+ }
+ // TODO : Support intermediate flush?
+ LOG.info("Closing the temp rename file {}", renameTempPath);
+ fsDataOutputStream.close();
+ if (fsDis != null) {
+ LOG.debug("Closing the input stream");
+ fsDis.close();
+ }
+ } catch (IOException e) {
+ LOG.error(
+ "Unable to recover. Exception while trying to truncate the temp file {}",
+ tempFile);
+ // We cannot recover. This we can control if user does not want this??
+ throw e;
+ }
+ try {
+ LOG.info("Deleting the actual temp file {}", tempFile);
+ fs.delete(tempFile, false);
+ } catch (IOException e) {
+ LOG.error("Unable to recover. Error while deleting the temp file {}", tempFile);
+ // unable to recover.
+ throw e;
+ }
+ rename(fs, renameTempPath);
Review Comment:
What if there is a fail in between these steps of delete and rename? Later how it will be possible to rollback. The truncate should be able to handle such case?
##########
flink-filesystems/flink-azure-fs-hadoop/src/main/java/org/apache/flink/fs/azurefs/AzureBlobFsRecoverableDataOutputStream.java:
##########
@@ -0,0 +1,299 @@
+/*
+ * 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.flink.fs.azurefs;
+
+import org.apache.flink.annotation.Internal;
+import org.apache.flink.core.fs.RecoverableFsDataOutputStream;
+import org.apache.flink.core.fs.RecoverableWriter;
+import org.apache.flink.runtime.fs.hdfs.HadoopFsRecoverable;
+
+import org.apache.hadoop.fs.FSDataInputStream;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.FileNotFoundException;
+import java.io.IOException;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+/**
+ * An implementation of the {@link RecoverableFsDataOutputStream} for AzureBlob's file system
+ * abstraction.
+ */
+@Internal
+public class AzureBlobFsRecoverableDataOutputStream extends RecoverableFsDataOutputStream {
+
+ private static final Logger LOG =
+ LoggerFactory.getLogger(AzureBlobFsRecoverableDataOutputStream.class);
+
+ private final FileSystem fs;
+
+ private final Path targetFile;
+
+ private final Path tempFile;
+
+ private final FSDataOutputStream out;
+
+ // Not final to override in tests
+ public static int minBufferLength = 2097152;
+
+ // init to 0. When ever recovery is done add this to the pos.
+ private long initialFileSize = 0;
+
+ public AzureBlobFsRecoverableDataOutputStream(FileSystem fs, Path targetFile, Path tempFile)
+ throws IOException {
+ this.fs = checkNotNull(fs);
+ this.targetFile = checkNotNull(targetFile);
+ LOG.debug("The targetFile is {}", targetFile.getName());
+ this.tempFile = checkNotNull(tempFile);
+ LOG.debug("The tempFile is {}", tempFile.getName());
+ this.out = fs.create(tempFile);
+ }
+
+ public AzureBlobFsRecoverableDataOutputStream(FileSystem fs, HadoopFsRecoverable recoverable)
+ throws IOException {
+ this.fs = checkNotNull(fs);
+ this.targetFile = checkNotNull(recoverable.targetFile());
+ this.tempFile = checkNotNull(recoverable.tempFile());
+ long len = fs.getFileStatus(tempFile).getLen();
+ LOG.info("The recoverable offset is {} and the file len is {}", recoverable.offset(), len);
+ // Happens when we recover from a previously committed offset. Otherwise this is not
+ // really needed
+ if (len > recoverable.offset()) {
+ truncate(fs, recoverable);
+ } else if (len < recoverable.offset()) {
+ LOG.error(
+ "Temp file length {} is less than the expected recoverable offset {}",
+ len,
+ recoverable.offset());
+ throw new IOException(
+ "Unable to create recoverable outputstream as length of file "
+ + len
+ + " is less than "
+ + "recoverable offset "
+ + recoverable.offset());
+ }
+ // In ABFS when we try to append we don't account for the initial file size like we do in
+ // DFS.
+ // So we explicitly store this and when we do a persist call we make use of it.
+ initialFileSize = fs.getFileStatus(tempFile).getLen();
+ out = fs.append(tempFile);
+ LOG.debug("Created a new OS for appending {}", tempFile);
+ }
+
+ private void truncate(FileSystem fs, HadoopFsRecoverable recoverable) throws IOException {
+ Path renameTempPath = new Path(tempFile.toString() + ".rename");
+ try {
+ LOG.info(
+ "Creating the temp rename file {} for truncating the tempFile {}",
+ renameTempPath,
+ tempFile);
+ FSDataOutputStream fsDataOutputStream = fs.create(renameTempPath);
Review Comment:
What if an attempt of recover failed itself in between? That will be retried again later and so a similar another truncate attempt? If so, we might have to do a cleanup of *.rename temp file itself. May be its ok as default FileSystem behave is overwrite.
--
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: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org