Examples of RandomAccessReader


Examples of org.apache.cassandra.io.util.RandomAccessReader

        // loop through each row, deserializing to check for damage.
        // we'll also loop through the index at the same time, using the position from the index to recover if the
        // row header (key or data size) is corrupt. (This means our position in the index file will be one row
        // "ahead" of the data file.)
        final RandomAccessReader dataFile = sstable.openDataReader(true);
        RandomAccessReader indexFile = RandomAccessReader.open(new File(sstable.descriptor.filenameFor(Component.PRIMARY_INDEX)), true);
        ScrubInfo scrubInfo = new ScrubInfo(dataFile, sstable);
        executor.beginCompaction(scrubInfo);

        SSTableWriter writer = null;
        SSTableReader newSstable = null;
        int goodRows = 0, badRows = 0, emptyRows = 0;

        try
        {
            ByteBuffer nextIndexKey = ByteBufferUtil.readWithShortLength(indexFile);
            {
                // throw away variable so we don't have a side effect in the assert
                long firstRowPositionFromIndex = indexFile.readLong();
                assert firstRowPositionFromIndex == 0 : firstRowPositionFromIndex;
            }

            // TODO errors when creating the writer may leave empty temp files.
            writer = maybeCreateWriter(cfs, compactionFileLocation, expectedBloomFilterSize, null, Collections.singletonList(sstable));

            while (!dataFile.isEOF())
            {
                long rowStart = dataFile.getFilePointer();
                if (logger.isDebugEnabled())
                    logger.debug("Reading row at " + rowStart);

                DecoratedKey key = null;
                long dataSize = -1;
                try
                {
                    key = SSTableReader.decodeKey(sstable.partitioner, sstable.descriptor, ByteBufferUtil.readWithShortLength(dataFile));
                    dataSize = sstable.descriptor.hasIntRowSize ? dataFile.readInt() : dataFile.readLong();
                    if (logger.isDebugEnabled())
                        logger.debug(String.format("row %s is %s bytes", ByteBufferUtil.bytesToHex(key.key), dataSize));
                }
                catch (Throwable th)
                {
                    throwIfFatal(th);
                    // check for null key below
                }

                ByteBuffer currentIndexKey = nextIndexKey;
                long nextRowPositionFromIndex;
                try
                {
                    nextIndexKey = indexFile.isEOF() ? null : ByteBufferUtil.readWithShortLength(indexFile);
                    nextRowPositionFromIndex = indexFile.isEOF() ? dataFile.length() : indexFile.readLong();
                }
                catch (Throwable th)
                {
                    logger.warn("Error reading index file", th);
                    nextIndexKey = null;
                    nextRowPositionFromIndex = dataFile.length();
                }

                long dataStart = dataFile.getFilePointer();
                long dataStartFromIndex = currentIndexKey == null
                                        ? -1
                                        : rowStart + 2 + currentIndexKey.remaining() + (sstable.descriptor.hasIntRowSize ? 4 : 8);
                long dataSizeFromIndex = nextRowPositionFromIndex - dataStartFromIndex;
                assert currentIndexKey != null || indexFile.isEOF();
                if (logger.isDebugEnabled() && currentIndexKey != null)
                    logger.debug(String.format("Index doublecheck: row %s is %s bytes", ByteBufferUtil.bytesToHex(currentIndexKey),  dataSizeFromIndex));

                writer.mark();
                try
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

        logger.info("Replaying " + file.getPath());
        CommitLogDescriptor desc = CommitLogDescriptor.fromFileName(file.getName());
        final long segment = desc.id;
        int version = desc.getMessagingVersion();
        RandomAccessReader reader = RandomAccessReader.open(new File(file.getAbsolutePath()));
        try
        {
            assert reader.length() <= Integer.MAX_VALUE;
            int replayPosition;
            if (globalPosition.segment < segment)
            {
                replayPosition = 0;
            }
            else if (globalPosition.segment == segment)
            {
                replayPosition = globalPosition.position;
            }
            else
            {
                logger.debug("skipping replay of fully-flushed {}", file);
                return;
            }

            if (logger.isDebugEnabled())
                logger.debug("Replaying " + file + " starting at " + replayPosition);
            reader.seek(replayPosition);

            /* read the logs populate RowMutation and apply */
            while (!reader.isEOF())
            {
                if (logger.isDebugEnabled())
                    logger.debug("Reading mutation at " + reader.getFilePointer());

                long claimedCRC32;
                int serializedSize;
                try
                {
                    // any of the reads may hit EOF
                    serializedSize = reader.readInt();
                    if (serializedSize == CommitLog.END_OF_SEGMENT_MARKER)
                    {
                        logger.debug("Encountered end of segment marker at " + reader.getFilePointer());
                        break;
                    }

                    // RowMutation must be at LEAST 10 bytes:
                    // 3 each for a non-empty Keyspace and Key (including the
                    // 2-byte length from writeUTF/writeWithShortLength) and 4 bytes for column count.
                    // This prevents CRC by being fooled by special-case garbage in the file; see CASSANDRA-2128
                    if (serializedSize < 10)
                        break;

                    long claimedSizeChecksum = reader.readLong();
                    checksum.reset();
                    if (version < CommitLogDescriptor.VERSION_20)
                        checksum.update(serializedSize);
                    else
                        FBUtilities.updateChecksumInt(checksum, serializedSize);

                    if (checksum.getValue() != claimedSizeChecksum)
                        break; // entry wasn't synced correctly/fully. that's
                               // ok.

                    if (serializedSize > buffer.length)
                        buffer = new byte[(int) (1.2 * serializedSize)];
                    reader.readFully(buffer, 0, serializedSize);
                    claimedCRC32 = reader.readLong();
                }
                catch (EOFException eof)
                {
                    break; // last CL entry didn't get completely written. that's ok.
                }

                checksum.update(buffer, 0, serializedSize);
                if (claimedCRC32 != checksum.getValue())
                {
                    // this entry must not have been fsynced. probably the rest is bad too,
                    // but just in case there is no harm in trying them (since we still read on an entry boundary)
                    continue;
                }

                /* deserialize the commit log entry */
                FastByteArrayInputStream bufIn = new FastByteArrayInputStream(buffer, 0, serializedSize);
                final RowMutation rm;
                try
                {
                    // assuming version here. We've gone to lengths to make sure what gets written to the CL is in
                    // the current version. so do make sure the CL is drained prior to upgrading a node.
                    rm = RowMutation.serializer.deserialize(new DataInputStream(bufIn), version, ColumnSerializer.Flag.LOCAL);
                    // doublecheck that what we read is [still] valid for the current schema
                    for (ColumnFamily cf : rm.getColumnFamilies())
                        for (Column cell : cf)
                            cf.getComparator().validate(cell.name());
                }
                catch (UnknownColumnFamilyException ex)
                {
                    if (ex.cfId == null)
                        continue;
                    AtomicInteger i = invalidMutations.get(ex.cfId);
                    if (i == null)
                    {
                        i = new AtomicInteger(1);
                        invalidMutations.put(ex.cfId, i);
                    }
                    else
                        i.incrementAndGet();
                    continue;
                }
                catch (Throwable t)
                {
                    File f = File.createTempFile("mutation", "dat");
                    DataOutputStream out = new DataOutputStream(new FileOutputStream(f));
                    try
                    {
                        out.write(buffer, 0, serializedSize);
                    }
                    finally
                    {
                        out.close();
                    }
                    String st = String.format("Unexpected error deserializing mutation; saved to %s and ignored.  This may be caused by replaying a mutation against a table with the same name but incompatible schema.  Exception follows: ",
                                              f.getAbsolutePath());
                    logger.error(st, t);
                    continue;
                }

                if (logger.isDebugEnabled())
                    logger.debug(String.format("replaying mutation for %s.%s: %s", rm.getKeyspaceName(), ByteBufferUtil.bytesToHex(rm.key()), "{" + StringUtils.join(rm.getColumnFamilies().iterator(), ", ")
                            + "}"));

                final long entryLocation = reader.getFilePointer();
                Runnable runnable = new WrappedRunnable()
                {
                    public void runMayThrow() throws IOException
                    {
                        if (Schema.instance.getKSMetaData(rm.getKeyspaceName()) == null)
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

    public RandomAccessReader get(String path)
    {
        metrics.requests.mark();

        Queue<RandomAccessReader> instances = getCacheFor(path);
        RandomAccessReader result = instances.poll();
        if (result != null)
            metrics.hits.mark();

        return result;
    }
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

     * @throws IOException on failure to read/write input/output
     */
    public static void export(Descriptor desc, PrintStream outs, Collection<String> toExport, String[] excludes) throws IOException
    {
        SSTableReader sstable = SSTableReader.open(desc);
        RandomAccessReader dfile = sstable.openDataReader();

        IPartitioner<?> partitioner = sstable.partitioner;

        if (excludes != null)
            toExport.removeAll(Arrays.asList(excludes));

        outs.println("[");

        int i = 0;

        // last key to compare order
        DecoratedKey lastKey = null;

        for (String key : toExport)
        {
            DecoratedKey decoratedKey = partitioner.decorateKey(hexToBytes(key));

            if (lastKey != null && lastKey.compareTo(decoratedKey) > 0)
                throw new IOException("Key out of order! " + lastKey + " > " + decoratedKey);

            lastKey = decoratedKey;

            RowIndexEntry entry = sstable.getPosition(decoratedKey, SSTableReader.Operator.EQ);
            if (entry == null)
                continue;

            dfile.seek(entry.position);
            ByteBufferUtil.readWithShortLength(dfile); // row key
            if (sstable.descriptor.version.hasRowSizeAndColumnCount)
                dfile.readLong(); // row size
            DeletionInfo deletionInfo = new DeletionInfo(DeletionTime.serializer.deserialize(dfile));
            int columnCount = sstable.descriptor.version.hasRowSizeAndColumnCount ? dfile.readInt() : Integer.MAX_VALUE;

            Iterator<OnDiskAtom> atomIterator = sstable.metadata.getOnDiskIterator(dfile, columnCount, sstable.descriptor.version);

            checkStream(outs);

View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

     * @throws IOException on any I/O error
     */
    public void write(WritableByteChannel channel) throws IOException
    {
        long totalSize = totalSize();
        RandomAccessReader file = sstable.openDataReader();
        ChecksumValidator validator = null;
        if (new File(sstable.descriptor.filenameFor(Component.CRC)).exists())
            validator = DataIntegrityMetadata.checksumValidator(sstable.descriptor);

        transferBuffer = validator == null ? new byte[DEFAULT_CHUNK_SIZE] : new byte[validator.chunkSize];

        // setting up data compression stream
        compressedOutput = new LZFOutputStream(Channels.newOutputStream(channel));
        long progress = 0L;

        try
        {
            // stream each of the required sections of the file
            for (Pair<Long, Long> section : sections)
            {
                long start = validator == null ? section.left : validator.chunkStart(section.left);
                int skipBytes = (int) (section.left - start);
                // seek to the beginning of the section
                file.seek(start);
                if (validator != null)
                    validator.seek(start);

                // length of the section to read
                long length = section.right - start;
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

        // loop through each row, deserializing to check for damage.
        // we'll also loop through the index at the same time, using the position from the index to recover if the
        // row header (key or data size) is corrupt. (This means our position in the index file will be one row
        // "ahead" of the data file.)
        final RandomAccessReader dataFile = sstable.openDataReader(true);
        RandomAccessReader indexFile = RandomAccessReader.open(new File(sstable.descriptor.filenameFor(Component.PRIMARY_INDEX)), true);
        ScrubInfo scrubInfo = new ScrubInfo(dataFile, sstable);
        executor.beginCompaction(scrubInfo);

        SSTableWriter writer = null;
        SSTableReader newSstable = null;
        int goodRows = 0, badRows = 0, emptyRows = 0;

        try
        {
            ByteBuffer nextIndexKey = ByteBufferUtil.readWithShortLength(indexFile);
            {
                // throw away variable so we don't have a side effect in the assert
                long firstRowPositionFromIndex = indexFile.readLong();
                assert firstRowPositionFromIndex == 0 : firstRowPositionFromIndex;
            }

            // TODO errors when creating the writer may leave empty temp files.
            writer = maybeCreateWriter(cfs, compactionFileLocation, expectedBloomFilterSize, null, Collections.singletonList(sstable));

            while (!dataFile.isEOF())
            {
                long rowStart = dataFile.getFilePointer();
                if (logger.isDebugEnabled())
                    logger.debug("Reading row at " + rowStart);

                DecoratedKey key = null;
                long dataSize = -1;
                try
                {
                    key = SSTableReader.decodeKey(sstable.partitioner, sstable.descriptor, ByteBufferUtil.readWithShortLength(dataFile));
                    dataSize = sstable.descriptor.hasIntRowSize ? dataFile.readInt() : dataFile.readLong();
                    if (logger.isDebugEnabled())
                        logger.debug(String.format("row %s is %s bytes", ByteBufferUtil.bytesToHex(key.key), dataSize));
                }
                catch (Throwable th)
                {
                    throwIfFatal(th);
                    // check for null key below
                }

                ByteBuffer currentIndexKey = nextIndexKey;
                long nextRowPositionFromIndex;
                try
                {
                    nextIndexKey = indexFile.isEOF() ? null : ByteBufferUtil.readWithShortLength(indexFile);
                    nextRowPositionFromIndex = indexFile.isEOF() ? dataFile.length() : indexFile.readLong();
                }
                catch (Throwable th)
                {
                    logger.warn("Error reading index file", th);
                    nextIndexKey = null;
                    nextRowPositionFromIndex = dataFile.length();
                }

                long dataStart = dataFile.getFilePointer();
                long dataStartFromIndex = currentIndexKey == null
                                        ? -1
                                        : rowStart + 2 + currentIndexKey.remaining() + (sstable.descriptor.hasIntRowSize ? 4 : 8);
                long dataSizeFromIndex = nextRowPositionFromIndex - dataStartFromIndex;
                assert currentIndexKey != null || indexFile.isEOF();
                if (logger.isDebugEnabled() && currentIndexKey != null)
                    logger.debug(String.format("Index doublecheck: row %s is %s bytes", ByteBufferUtil.bytesToHex(currentIndexKey),  dataSizeFromIndex));

                writer.mark();
                try
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

            CompactionManager.instance.performMaximal(cfStore);
        }
        // verify that we do indeed have multiple index entries
        SSTableReader sstable = cfStore.getSSTables().iterator().next();
        long position = sstable.getPosition(key, SSTableReader.Operator.EQ);
        RandomAccessReader file = sstable.openDataReader(false);
        file.seek(position);
        assert ByteBufferUtil.readWithShortLength(file).equals(key.key);
        SSTableReader.readRowSize(file, sstable.descriptor);
        IndexHelper.skipBloomFilter(file);
        ArrayList<IndexHelper.IndexInfo> indexes = IndexHelper.deserializeIndex(file);
        assert indexes.size() > 2;
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

        // loop through each row, deserializing to check for damage.
        // we'll also loop through the index at the same time, using the position from the index to recover if the
        // row header (key or data size) is corrupt. (This means our position in the index file will be one row
        // "ahead" of the data file.)
        final RandomAccessReader dataFile = sstable.openDataReader(true);

        String indexFilename = sstable.descriptor.filenameFor(Component.PRIMARY_INDEX);
        RandomAccessReader indexFile = RandomAccessReader.open(new File(indexFilename), true);
        try
        {
            ByteBuffer nextIndexKey = ByteBufferUtil.readWithShortLength(indexFile);
            {
                // throw away variable so we don't have a side effect in the assert
                long firstRowPositionFromIndex = indexFile.readLong();
                assert firstRowPositionFromIndex == 0 : firstRowPositionFromIndex;
            }

            // errors when creating the writer may leave empty temp files.
            SSTableWriter writer = maybeCreateWriter(cfs, compactionFileLocation, expectedBloomFilterSize, null, Collections.singletonList(sstable));
            SSTableReader newSstable = null;
            executor.beginCompaction(new ScrubInfo(dataFile, sstable));
            int goodRows = 0, badRows = 0, emptyRows = 0;

            try
            {
                while (!dataFile.isEOF())
                {
                    long rowStart = dataFile.getFilePointer();
                    if (logger.isDebugEnabled())
                        logger.debug("Reading row at " + rowStart);

                    DecoratedKey key = null;
                    long dataSize = -1;
                    try
                    {
                        key = SSTableReader.decodeKey(sstable.partitioner, sstable.descriptor, ByteBufferUtil.readWithShortLength(dataFile));
                        dataSize = sstable.descriptor.hasIntRowSize ? dataFile.readInt() : dataFile.readLong();
                        if (logger.isDebugEnabled())
                            logger.debug(String.format("row %s is %s bytes", ByteBufferUtil.bytesToHex(key.key), dataSize));
                    }
                    catch (Throwable th)
                    {
                        throwIfFatal(th);
                        // check for null key below
                    }

                    ByteBuffer currentIndexKey = nextIndexKey;
                    long nextRowPositionFromIndex;
                    try
                    {
                        nextIndexKey = indexFile.isEOF() ? null : ByteBufferUtil.readWithShortLength(indexFile);
                        nextRowPositionFromIndex = indexFile.isEOF() ? dataFile.length() : indexFile.readLong();
                    }
                    catch (Throwable th)
                    {
                        logger.warn("Error reading index file", th);
                        nextIndexKey = null;
                        nextRowPositionFromIndex = dataFile.length();
                    }

                    long dataStart = dataFile.getFilePointer();
                    long dataStartFromIndex = currentIndexKey == null
                                            ? -1
                                            : rowStart + 2 + currentIndexKey.remaining() + (sstable.descriptor.hasIntRowSize ? 4 : 8);
                    long dataSizeFromIndex = nextRowPositionFromIndex - dataStartFromIndex;
                    assert currentIndexKey != null || indexFile.isEOF();
                    if (logger.isDebugEnabled() && currentIndexKey != null)
                        logger.debug(String.format("Index doublecheck: row %s is %s bytes", ByteBufferUtil.bytesToHex(currentIndexKey),  dataSizeFromIndex));

                    writer.mark();
                    try
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

        if (header.file == null)
            return;

        // TODO just use a raw RandomAccessFile since we're managing our own buffer here
        RandomAccessReader file = (header.file.sstable.compression) // try to skip kernel page cache if possible
                                ? CompressedRandomAccessReader.open(header.file.getFilename(), true)
                                : RandomAccessReader.open(new File(header.file.getFilename()), true);

        // setting up data compression stream
        output = new LZFOutputStream(output);

        try
        {
            // stream each of the required sections of the file
            for (Pair<Long, Long> section : header.file.sections)
            {
                // seek to the beginning of the section
                file.seek(section.left);

                // length of the section to stream
                long length = section.right - section.left;
                // tracks write progress
                long bytesTransferred = 0;
View Full Code Here

Examples of org.apache.cassandra.io.util.RandomAccessReader

        try
        {
            if (input instanceof RandomAccessReader)
            {
                RandomAccessReader file = (RandomAccessReader) input;
                file.seek(this.dataStart);
                if (dataStart + dataSize > file.length())
                    throw new IOException(String.format("dataSize of %s starting at %s would be larger than file %s length %s",
                                          dataSize, dataStart, file.getPath(), file.length()));
                if (checkData)
                {
                    try
                    {
                        IndexHelper.defreezeBloomFilter(file, dataSize, sstable.descriptor.usesOldBloomFilter);
                    }
                    catch (Exception e)
                    {
                        if (e instanceof EOFException)
                            throw (EOFException) e;

                        logger.debug("Invalid bloom filter in {}; will rebuild it", sstable);
                        // deFreeze should have left the file position ready to deserialize index
                    }
                    try
                    {
                        IndexHelper.deserializeIndex(file);
                    }
                    catch (Exception e)
                    {
                        logger.debug("Invalid row summary in {}; will rebuild it", sstable);
                    }
                    file.seek(this.dataStart);
                    inputWithTracker.reset(0);
                }
            }

            IndexHelper.skipBloomFilter(inputWithTracker);
View Full Code Here
TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.