HBASE-27727 Implement filesystem based Replication peer storage (#5165)

Signed-off-by: Liangjun He <heliangjun@apache.org>

hbase-client/src/main/java/org/apache/hadoop/hbase/replication/SyncReplicationState.java

@@ -17,10 +17,16 @@
  */
 package org.apache.hadoop.hbase.replication;
 
+import java.io.ByteArrayInputStream;
+import java.io.ByteArrayOutputStream;
+import java.io.IOException;
 import java.util.Arrays;
 import org.apache.hadoop.hbase.client.replication.ReplicationPeerConfigUtil;
+import org.apache.hadoop.hbase.protobuf.ProtobufMagic;
+import org.apache.hadoop.hbase.util.Pair;
 import org.apache.yetus.audience.InterfaceAudience;
 
+import org.apache.hbase.thirdparty.com.google.common.io.ByteStreams;
 import org.apache.hbase.thirdparty.com.google.protobuf.InvalidProtocolBufferException;
 
 import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil;
@@ -74,4 +80,28 @@ public static SyncReplicationState parseFrom(byte[] bytes) throws InvalidProtoco
     return ReplicationPeerConfigUtil.toSyncReplicationState(ReplicationProtos.SyncReplicationState
       .parseFrom(Arrays.copyOfRange(bytes, ProtobufUtil.lengthOfPBMagic(), bytes.length)));
   }
+
+  public static byte[] toByteArray(SyncReplicationState state, SyncReplicationState newState) {
+    ByteArrayOutputStream out = new ByteArrayOutputStream();
+    try {
+      out.write(ProtobufMagic.PB_MAGIC);
+      ReplicationPeerConfigUtil.toSyncReplicationState(state).writeDelimitedTo(out);
+      ReplicationPeerConfigUtil.toSyncReplicationState(newState).writeDelimitedTo(out);
+    } catch (IOException e) {
+      // should not happen, all in memory operations
+      throw new AssertionError(e);
+    }
+    return out.toByteArray();
+  }
+
+  public static Pair<SyncReplicationState, SyncReplicationState>
+    parseStateAndNewStateFrom(byte[] bytes) throws IOException {
+    ByteArrayInputStream in = new ByteArrayInputStream(bytes);
+    ByteStreams.skipFully(in, ProtobufMagic.lengthOfPBMagic());
+    SyncReplicationState state = ReplicationPeerConfigUtil
+      .toSyncReplicationState(ReplicationProtos.SyncReplicationState.parseDelimitedFrom(in));
+    SyncReplicationState newState = ReplicationPeerConfigUtil
+      .toSyncReplicationState(ReplicationProtos.SyncReplicationState.parseDelimitedFrom(in));
+    return Pair.newPair(state, newState);
+  }
 }

hbase-client/src/main/java/org/apache/hadoop/hbase/util/RotateFile.java

@@ -0,0 +1,197 @@
+/*
+ * 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.util;
+
+import com.google.errorprone.annotations.RestrictedApi;
+import java.io.EOFException;
+import java.io.FileNotFoundException;
+import java.io.IOException;
+import java.util.zip.CRC32;
+import org.apache.hadoop.fs.FSDataInputStream;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.apache.yetus.audience.InterfaceAudience;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.hbase.thirdparty.com.google.protobuf.ByteString;
+
+import org.apache.hadoop.hbase.shaded.protobuf.generated.HBaseProtos;
+
+/**
+ * A file storage which supports atomic update through two files, i.e, rotating. The implementation
+ * does not require atomic rename.
+ */
+@InterfaceAudience.Private
+public class RotateFile {
+
+  private static final Logger LOG = LoggerFactory.getLogger(RotateFile.class);
+
+  private final FileSystem fs;
+
+  private final long maxFileSize;
+
+  private final Path[] files = new Path[2];
+
+  // this is used to make sure that we do not go backwards
+  private long prevTimestamp = -1;
+
+  private int nextFile = -1;
+
+  /**
+   * Constructs a new RotateFile object with the given parameters.
+   * @param fs          the file system to use.
+   * @param dir         the directory where the files will be created.
+   * @param name        the base name for the files.
+   * @param maxFileSize the maximum size of each file.
+   */
+  public RotateFile(FileSystem fs, Path dir, String name, long maxFileSize) {
+    this.fs = fs;
+    this.maxFileSize = maxFileSize;
+    this.files[0] = new Path(dir, name + "-0");
+    this.files[1] = new Path(dir, name + "-1");
+  }
+
+  private HBaseProtos.RotateFileData read(Path path) throws IOException {
+    byte[] data;
+    int expectedChecksum;
+    try (FSDataInputStream in = fs.open(path)) {
+      int length = in.readInt();
+      if (length <= 0 || length > maxFileSize) {
+        throw new IOException("Invalid file length " + length
+          + ", either less than 0 or greater then max allowed size " + maxFileSize);
+      }
+      data = new byte[length];
+      in.readFully(data);
+      expectedChecksum = in.readInt();
+    }
+    CRC32 crc32 = new CRC32();
+    crc32.update(data);
+    int calculatedChecksum = (int) crc32.getValue();
+    if (expectedChecksum != calculatedChecksum) {
+      throw new IOException(
+        "Checksum mismatch, expected " + expectedChecksum + ", actual " + calculatedChecksum);
+    }
+    return HBaseProtos.RotateFileData.parseFrom(data);
+  }
+
+  private int select(HBaseProtos.RotateFileData[] datas) {
+    if (datas[0] == null) {
+      return 1;
+    }
+    if (datas[1] == null) {
+      return 0;
+    }
+    return datas[0].getTimestamp() >= datas[1].getTimestamp() ? 0 : 1;
+  }
+
+  /**
+   * Reads the content of the rotate file by selecting the winner file based on the timestamp of the
+   * data inside the files. It reads the content of both files and selects the one with the latest
+   * timestamp as the winner. If a file is incomplete or does not exist, it logs the error and moves
+   * on to the next file. It returns the content of the winner file as a byte array. If none of the
+   * files have valid data, it returns null.
+   * @return a byte array containing the data from the winner file, or null if no valid data is
+   *         found.
+   * @throws IOException if an error occurs while reading the files.
+   */
+  public byte[] read() throws IOException {
+    HBaseProtos.RotateFileData[] datas = new HBaseProtos.RotateFileData[2];
+    for (int i = 0; i < 2; i++) {
+      try {
+        datas[i] = read(files[i]);
+      } catch (FileNotFoundException e) {
+        LOG.debug("file {} does not exist", files[i], e);
+      } catch (EOFException e) {
+        LOG.debug("file {} is incomplete", files[i], e);
+      }
+    }
+    int winnerIndex = select(datas);
+    nextFile = 1 - winnerIndex;
+    if (datas[winnerIndex] != null) {
+      prevTimestamp = datas[winnerIndex].getTimestamp();
+      return datas[winnerIndex].getData().toByteArray();
+    } else {
+      return null;
+    }
+  }
+
+  @RestrictedApi(explanation = "Should only be called in tests", link = "",
+      allowedOnPath = ".*/RotateFile.java|.*/src/test/.*")
+  static void write(FileSystem fs, Path file, long timestamp, byte[] data) throws IOException {
+    HBaseProtos.RotateFileData proto = HBaseProtos.RotateFileData.newBuilder()
+      .setTimestamp(timestamp).setData(ByteString.copyFrom(data)).build();
+    byte[] protoData = proto.toByteArray();
+    CRC32 crc32 = new CRC32();
+    crc32.update(protoData);
+    int checksum = (int) crc32.getValue();
+    // 4 bytes length, 8 bytes timestamp, 4 bytes checksum at the end
+    try (FSDataOutputStream out = fs.create(file, true)) {
+      out.writeInt(protoData.length);
+      out.write(protoData);
+      out.writeInt(checksum);
+    }
+  }
+
+  /**
+   * Writes the given data to the next file in the rotation, with a timestamp calculated based on
+   * the previous timestamp and the current time to make sure it is greater than the previous
+   * timestamp. The method also deletes the previous file, which is no longer needed.
+   * <p/>
+   * Notice that, for a newly created {@link RotateFile} instance, you need to call {@link #read()}
+   * first to initialize the nextFile index, before calling this method.
+   * @param data the data to be written to the file
+   * @throws IOException if an I/O error occurs while writing the data to the file
+   */
+  public void write(byte[] data) throws IOException {
+    if (data.length > maxFileSize) {
+      throw new IOException(
+        "Data size " + data.length + " is greater than max allowed size " + maxFileSize);
+    }
+    long timestamp = Math.max(prevTimestamp + 1, EnvironmentEdgeManager.currentTime());
+    write(fs, files[nextFile], timestamp, data);
+    prevTimestamp = timestamp;
+    nextFile = 1 - nextFile;
+    try {
+      fs.delete(files[nextFile], false);
+    } catch (IOException e) {
+      // we will create new file with overwrite = true, so not a big deal here, only for speed up
+      // loading as we do not need to read this file when loading
+      LOG.debug("Failed to delete old file {}, ignoring the exception", files[nextFile], e);
+    }
+  }
+
+  /**
+   * Deletes the two files used for rotating data. If any of the files cannot be deleted, an
+   * IOException is thrown.
+   * @throws IOException if there is an error deleting either file
+   */
+  public void delete() throws IOException {
+    Path next = files[nextFile];
+    // delete next file first, and then the current file, so when failing to delete, we can still
+    // read the correct data
+    if (fs.exists(next) && !fs.delete(next, false)) {
+      throw new IOException("Can not delete " + next);
+    }
+    Path current = files[1 - nextFile];
+    if (fs.exists(current) && !fs.delete(current, false)) {
+      throw new IOException("Can not delete " + current);
+    }
+  }
+}