2 * Copyright 2017-present Open Networking Foundation
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
16 package io.atomix.storage.journal;
18 import com.esotericsoftware.kryo.KryoException;
19 import io.atomix.storage.journal.index.JournalIndex;
21 import java.io.IOException;
22 import java.nio.BufferOverflowException;
23 import java.nio.BufferUnderflowException;
24 import java.nio.ByteBuffer;
25 import java.nio.MappedByteBuffer;
26 import java.nio.channels.FileChannel;
27 import java.util.zip.CRC32;
28 import org.eclipse.jdt.annotation.NonNull;
33 * The format of an entry in the log is as follows:
35 * <li>64-bit index</li>
36 * <li>8-bit boolean indicating whether a term change is contained in the entry</li>
37 * <li>64-bit optional term</li>
38 * <li>32-bit signed entry length, including the entry type ID</li>
39 * <li>8-bit signed entry type ID</li>
40 * <li>n-bit entry bytes</li>
43 * @author <a href="http://github.com/kuujo">Jordan Halterman</a>
45 final class MappedJournalSegmentWriter<E> extends JournalSegmentWriter<E> {
46 private final @NonNull MappedByteBuffer mappedBuffer;
47 private final ByteBuffer buffer;
49 private Indexed<E> lastEntry;
51 MappedJournalSegmentWriter(
53 JournalSegment<E> segment,
56 JournalSerdes namespace) {
57 super(channel, segment, maxEntrySize, index, namespace);
59 mappedBuffer = channel.map(FileChannel.MapMode.READ_WRITE, 0, segment.descriptor().maxSegmentSize());
60 } catch (IOException e) {
61 throw new StorageException(e);
63 this.buffer = mappedBuffer.slice();
68 @NonNull MappedByteBuffer buffer() {
73 MappedJournalSegmentWriter<E> toMapped() {
78 FileChannelJournalSegmentWriter<E> toFileChannel() {
80 return new FileChannelJournalSegmentWriter<>(channel, segment, maxEntrySize, index, namespace);
84 public void reset(long index) {
85 long nextIndex = firstIndex;
87 // Clear the buffer indexes.
88 buffer.position(JournalSegmentDescriptor.BYTES);
90 // Record the current buffer position.
91 int position = buffer.position();
93 // Read the entry length.
97 int length = buffer.getInt();
99 // If the length is non-zero, read the entry.
100 while (0 < length && length <= maxEntrySize && (index == 0 || nextIndex <= index)) {
102 // Read the checksum of the entry.
103 final long checksum = buffer.getInt() & 0xFFFFFFFFL;
105 // Slice off the entry's bytes
106 final ByteBuffer entryBytes = buffer.slice();
107 entryBytes.limit(length);
109 // Compute the checksum for the entry bytes.
110 final CRC32 crc32 = new CRC32();
111 crc32.update(entryBytes);
113 // If the stored checksum does not equal the computed checksum, do not proceed further
114 if (checksum != crc32.getValue()) {
119 final E entry = namespace.deserialize(entryBytes);
120 lastEntry = new Indexed<>(nextIndex, entry, length);
121 this.index.index(nextIndex, position);
124 // Update the current position for indexing.
125 position = buffer.position() + length;
126 buffer.position(position);
129 length = buffer.getInt();
132 // Reset the buffer to the previous mark.
134 } catch (BufferUnderflowException e) {
140 public long getLastIndex() {
141 return lastEntry != null ? lastEntry.index() : segment.index() - 1;
145 public Indexed<E> getLastEntry() {
150 public long getNextIndex() {
151 if (lastEntry != null) {
152 return lastEntry.index() + 1;
159 public void append(Indexed<E> entry) {
160 final long nextIndex = getNextIndex();
162 // If the entry's index is greater than the next index in the segment, skip some entries.
163 if (entry.index() > nextIndex) {
164 throw new IndexOutOfBoundsException("Entry index is not sequential");
167 // If the entry's index is less than the next index, truncate the segment.
168 if (entry.index() < nextIndex) {
169 truncate(entry.index() - 1);
171 append(entry.entry());
175 @SuppressWarnings("unchecked")
176 public <T extends E> Indexed<T> append(T entry) {
177 // Store the entry index.
178 final long index = getNextIndex();
180 // Serialize the entry.
181 int position = buffer.position();
182 if (position + Integer.BYTES + Integer.BYTES > buffer.limit()) {
183 throw new BufferOverflowException();
186 buffer.position(position + Integer.BYTES + Integer.BYTES);
189 namespace.serialize(entry, buffer);
190 } catch (KryoException e) {
191 throw new BufferOverflowException();
194 final int length = buffer.position() - (position + Integer.BYTES + Integer.BYTES);
196 // If the entry length exceeds the maximum entry size then throw an exception.
197 if (length > maxEntrySize) {
198 // Just reset the buffer. There's no need to zero the bytes since we haven't written the length or checksum.
199 buffer.position(position);
200 throw new StorageException.TooLarge("Entry size " + length + " exceeds maximum allowed bytes (" + maxEntrySize + ")");
203 // Compute the checksum for the entry.
204 final CRC32 crc32 = new CRC32();
205 buffer.position(position + Integer.BYTES + Integer.BYTES);
206 ByteBuffer slice = buffer.slice();
209 final long checksum = crc32.getValue();
211 // Create a single byte[] in memory for the entire entry and write it as a batch to the underlying buffer.
212 buffer.position(position);
213 buffer.putInt(length);
214 buffer.putInt((int) checksum);
215 buffer.position(position + Integer.BYTES + Integer.BYTES + length);
217 // Update the last entry with the correct index/term/length.
218 Indexed<E> indexedEntry = new Indexed<>(index, entry, length);
219 this.lastEntry = indexedEntry;
220 this.index.index(index, position);
221 return (Indexed<T>) indexedEntry;
226 public void truncate(long index) {
227 // If the index is greater than or equal to the last index, skip the truncate.
228 if (index >= getLastIndex()) {
232 // Reset the last entry.
235 // Truncate the index.
236 this.index.truncate(index);
238 if (index < segment.index()) {
239 // Reset the writer to the first entry.
240 buffer.position(JournalSegmentDescriptor.BYTES);
242 // Reset the writer to the given index.
246 // Zero the entry header at current buffer position.
247 int position = buffer.position();
248 // Note: we issue a single putLong() instead of two putInt()s.
250 buffer.position(position);
254 public void flush() {
255 mappedBuffer.force();
259 public void close() {
262 BufferCleaner.freeBuffer(mappedBuffer);
263 } catch (IOException e) {
264 throw new StorageException(e);