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.channels.FileChannel;
26 import java.util.zip.CRC32;
31 * The format of an entry in the log is as follows:
33 * <li>64-bit index</li>
34 * <li>8-bit boolean indicating whether a term change is contained in the entry</li>
35 * <li>64-bit optional term</li>
36 * <li>32-bit signed entry length, including the entry type ID</li>
37 * <li>8-bit signed entry type ID</li>
38 * <li>n-bit entry bytes</li>
41 * @author <a href="http://github.com/kuujo">Jordan Halterman</a>
43 class FileChannelJournalSegmentWriter<E> implements JournalWriter<E> {
44 private static final ByteBuffer ZERO_ENTRY_HEADER = ByteBuffer.wrap(new byte[Integer.BYTES + Integer.BYTES]);
46 private final FileChannel channel;
47 private final JournalSegment<E> segment;
48 private final int maxEntrySize;
49 private final JournalIndex index;
50 private final JournalSerdes namespace;
51 private final ByteBuffer memory;
52 private final long firstIndex;
53 private Indexed<E> lastEntry;
55 FileChannelJournalSegmentWriter(
57 JournalSegment<E> segment,
60 JournalSerdes namespace) {
61 this.channel = channel;
62 this.segment = segment;
63 this.maxEntrySize = maxEntrySize;
65 this.memory = ByteBuffer.allocate((maxEntrySize + Integer.BYTES + Integer.BYTES) * 2);
67 this.namespace = namespace;
68 this.firstIndex = segment.index();
73 public void reset(long index) {
74 long nextIndex = firstIndex;
76 // Clear the buffer indexes.
78 channel.position(JournalSegmentDescriptor.BYTES);
80 // Record the current buffer position.
81 long position = channel.position();
83 // Clear memory buffer and read fist chunk
85 channel.read(memory, position);
88 // Read the entry length.
90 int length = memory.getInt();
92 // If the length is non-zero, read the entry.
93 while (0 < length && length <= maxEntrySize && (index == 0 || nextIndex <= index)) {
95 // Read the checksum of the entry.
96 final long checksum = memory.getInt() & 0xFFFFFFFFL;
98 // Compute the checksum for the entry bytes.
99 final CRC32 crc32 = new CRC32();
100 crc32.update(memory.array(), memory.position(), length);
102 // If the stored checksum does not equal the computed checksum, do not proceed further
103 if (checksum != crc32.getValue()) {
107 int limit = memory.limit();
108 memory.limit(memory.position() + length);
109 final E entry = namespace.deserialize(memory);
111 lastEntry = new Indexed<>(nextIndex, entry, length);
112 this.index.index(nextIndex, (int) position);
115 // Update the current position for indexing.
116 position = channel.position() + memory.position();
118 // Read more bytes from the segment if necessary.
119 if (memory.remaining() < maxEntrySize) {
121 channel.position(position);
122 channel.read(memory, position);
127 length = memory.getInt();
130 // Reset the buffer to the previous mark.
131 channel.position(channel.position() + memory.reset().position());
132 } catch (BufferUnderflowException e) {
134 channel.position(channel.position() + memory.reset().position());
135 } catch (IOException e2) {
136 throw new StorageException(e2);
138 } catch (IOException e) {
139 throw new StorageException(e);
144 public long getLastIndex() {
145 return lastEntry != null ? lastEntry.index() : segment.index() - 1;
149 public Indexed<E> getLastEntry() {
154 public long getNextIndex() {
155 if (lastEntry != null) {
156 return lastEntry.index() + 1;
163 public void append(Indexed<E> entry) {
164 final long nextIndex = getNextIndex();
166 // If the entry's index is greater than the next index in the segment, skip some entries.
167 if (entry.index() > nextIndex) {
168 throw new IndexOutOfBoundsException("Entry index is not sequential");
171 // If the entry's index is less than the next index, truncate the segment.
172 if (entry.index() < nextIndex) {
173 truncate(entry.index() - 1);
175 append(entry.entry());
179 @SuppressWarnings("unchecked")
180 public <T extends E> Indexed<T> append(T entry) {
181 // Store the entry index.
182 final long index = getNextIndex();
185 // Serialize the entry.
187 memory.position(Integer.BYTES + Integer.BYTES);
189 namespace.serialize(entry, memory);
190 } catch (KryoException e) {
191 throw new StorageException.TooLarge("Entry size exceeds maximum allowed bytes (" + maxEntrySize + ")");
195 final int length = memory.limit() - (Integer.BYTES + Integer.BYTES);
197 // Ensure there's enough space left in the buffer to store the entry.
198 long position = channel.position();
199 if (segment.descriptor().maxSegmentSize() - position < length + Integer.BYTES + Integer.BYTES) {
200 throw new BufferOverflowException();
203 // If the entry length exceeds the maximum entry size then throw an exception.
204 if (length > maxEntrySize) {
205 throw new StorageException.TooLarge("Entry size " + length + " exceeds maximum allowed bytes (" + maxEntrySize + ")");
208 // Compute the checksum for the entry.
209 final CRC32 crc32 = new CRC32();
210 crc32.update(memory.array(), Integer.BYTES + Integer.BYTES, memory.limit() - (Integer.BYTES + Integer.BYTES));
211 final long checksum = crc32.getValue();
213 // Create a single byte[] in memory for the entire entry and write it as a batch to the underlying buffer.
214 memory.putInt(0, length);
215 memory.putInt(Integer.BYTES, (int) checksum);
216 channel.write(memory);
218 // Update the last entry with the correct index/term/length.
219 Indexed<E> indexedEntry = new Indexed<>(index, entry, length);
220 this.lastEntry = indexedEntry;
221 this.index.index(index, (int) position);
222 return (Indexed<T>) indexedEntry;
223 } catch (IOException e) {
224 throw new StorageException(e);
229 public void commit(long index) {
234 public void truncate(long index) {
235 // If the index is greater than or equal to the last index, skip the truncate.
236 if (index >= getLastIndex()) {
240 // Reset the last entry.
243 // Truncate the index.
244 this.index.truncate(index);
247 if (index < segment.index()) {
248 // Reset the writer to the first entry.
249 channel.position(JournalSegmentDescriptor.BYTES);
251 // Reset the writer to the given index.
255 // Zero the entry header at current channel position.
256 channel.write(ZERO_ENTRY_HEADER.asReadOnlyBuffer(), channel.position());
257 } catch (IOException e) {
258 throw new StorageException(e);
263 public void flush() {
265 if (channel.isOpen()) {
268 } catch (IOException e) {
269 throw new StorageException(e);
274 public void close() {