Search in sources :

Example 51 with HFileContext

use of org.apache.hadoop.hbase.io.hfile.HFileContext in project hbase by apache.

the class TestStoreFile method testBasicHalfMapFile.

/**
   * Write a file and then assert that we can read from top and bottom halves
   * using two HalfMapFiles.
   * @throws Exception
   */
@Test
public void testBasicHalfMapFile() throws Exception {
    final HRegionInfo hri = new HRegionInfo(TableName.valueOf("testBasicHalfMapFileTb"));
    HRegionFileSystem regionFs = HRegionFileSystem.createRegionOnFileSystem(conf, fs, new Path(testDir, hri.getTable().getNameAsString()), hri);
    HFileContext meta = new HFileContextBuilder().withBlockSize(2 * 1024).build();
    StoreFileWriter writer = new StoreFileWriter.Builder(conf, cacheConf, this.fs).withFilePath(regionFs.createTempName()).withFileContext(meta).build();
    writeStoreFile(writer);
    Path sfPath = regionFs.commitStoreFile(TEST_FAMILY, writer.getPath());
    StoreFile sf = new StoreFile(this.fs, sfPath, conf, cacheConf, BloomType.NONE);
    checkHalfHFile(regionFs, sf);
}
Also used : HRegionInfo(org.apache.hadoop.hbase.HRegionInfo) Path(org.apache.hadoop.fs.Path) HFileContextBuilder(org.apache.hadoop.hbase.io.hfile.HFileContextBuilder) HFileContext(org.apache.hadoop.hbase.io.hfile.HFileContext) Test(org.junit.Test)

Example 52 with HFileContext

use of org.apache.hadoop.hbase.io.hfile.HFileContext in project hbase by apache.

the class TestStoreFile method testBloomFilter.

@Test
public void testBloomFilter() throws Exception {
    FileSystem fs = FileSystem.getLocal(conf);
    conf.setFloat(BloomFilterFactory.IO_STOREFILE_BLOOM_ERROR_RATE, (float) 0.01);
    conf.setBoolean(BloomFilterFactory.IO_STOREFILE_BLOOM_ENABLED, true);
    // write the file
    Path f = new Path(ROOT_DIR, getName());
    HFileContext meta = new HFileContextBuilder().withBlockSize(BLOCKSIZE_SMALL).withChecksumType(CKTYPE).withBytesPerCheckSum(CKBYTES).build();
    // Make a store file and write data to it.
    StoreFileWriter writer = new StoreFileWriter.Builder(conf, cacheConf, this.fs).withFilePath(f).withBloomType(BloomType.ROW).withMaxKeyCount(2000).withFileContext(meta).build();
    bloomWriteRead(writer, fs);
}
Also used : Path(org.apache.hadoop.fs.Path) FileSystem(org.apache.hadoop.fs.FileSystem) HFileContextBuilder(org.apache.hadoop.hbase.io.hfile.HFileContextBuilder) HFileContext(org.apache.hadoop.hbase.io.hfile.HFileContext) Test(org.junit.Test)

Example 53 with HFileContext

use of org.apache.hadoop.hbase.io.hfile.HFileContext in project hbase by apache.

the class TestStoreFile method testBloomTypes.

@Test
public void testBloomTypes() throws Exception {
    float err = (float) 0.01;
    FileSystem fs = FileSystem.getLocal(conf);
    conf.setFloat(BloomFilterFactory.IO_STOREFILE_BLOOM_ERROR_RATE, err);
    conf.setBoolean(BloomFilterFactory.IO_STOREFILE_BLOOM_ENABLED, true);
    int rowCount = 50;
    int colCount = 10;
    int versions = 2;
    // run once using columns and once using rows
    BloomType[] bt = { BloomType.ROWCOL, BloomType.ROW };
    int[] expKeys = { rowCount * colCount, rowCount };
    // below line deserves commentary.  it is expected bloom false positives
    //  column = rowCount*2*colCount inserts
    //  row-level = only rowCount*2 inserts, but failures will be magnified by
    //              2nd for loop for every column (2*colCount)
    float[] expErr = { 2 * rowCount * colCount * err, 2 * rowCount * 2 * colCount * err };
    for (int x : new int[] { 0, 1 }) {
        // write the file
        Path f = new Path(ROOT_DIR, getName() + x);
        HFileContext meta = new HFileContextBuilder().withBlockSize(BLOCKSIZE_SMALL).withChecksumType(CKTYPE).withBytesPerCheckSum(CKBYTES).build();
        // Make a store file and write data to it.
        StoreFileWriter writer = new StoreFileWriter.Builder(conf, cacheConf, this.fs).withFilePath(f).withBloomType(bt[x]).withMaxKeyCount(expKeys[x]).withFileContext(meta).build();
        long now = System.currentTimeMillis();
        for (int i = 0; i < rowCount * 2; i += 2) {
            // rows
            for (int j = 0; j < colCount * 2; j += 2) {
                // column qualifiers
                String row = String.format(localFormatter, i);
                String col = String.format(localFormatter, j);
                for (int k = 0; k < versions; ++k) {
                    // versions
                    KeyValue kv = new KeyValue(row.getBytes(), "family".getBytes(), ("col" + col).getBytes(), now - k, Bytes.toBytes((long) -1));
                    writer.append(kv);
                }
            }
        }
        writer.close();
        StoreFileReader reader = new StoreFileReader(fs, f, cacheConf, conf);
        reader.loadFileInfo();
        reader.loadBloomfilter();
        StoreFileScanner scanner = getStoreFileScanner(reader, false, false);
        assertEquals(expKeys[x], reader.generalBloomFilter.getKeyCount());
        Store store = mock(Store.class);
        HColumnDescriptor hcd = mock(HColumnDescriptor.class);
        when(hcd.getName()).thenReturn(Bytes.toBytes("family"));
        when(store.getFamily()).thenReturn(hcd);
        // check false positives rate
        int falsePos = 0;
        int falseNeg = 0;
        for (int i = 0; i < rowCount * 2; ++i) {
            // rows
            for (int j = 0; j < colCount * 2; ++j) {
                // column qualifiers
                String row = String.format(localFormatter, i);
                String col = String.format(localFormatter, j);
                TreeSet<byte[]> columns = new TreeSet<>(Bytes.BYTES_COMPARATOR);
                columns.add(("col" + col).getBytes());
                Scan scan = new Scan(row.getBytes(), row.getBytes());
                scan.addColumn("family".getBytes(), ("col" + col).getBytes());
                boolean exists = scanner.shouldUseScanner(scan, store, Long.MIN_VALUE);
                boolean shouldRowExist = i % 2 == 0;
                boolean shouldColExist = j % 2 == 0;
                shouldColExist = shouldColExist || bt[x] == BloomType.ROW;
                if (shouldRowExist && shouldColExist) {
                    if (!exists)
                        falseNeg++;
                } else {
                    if (exists)
                        falsePos++;
                }
            }
        }
        // evict because we are about to delete the file
        reader.close(true);
        fs.delete(f, true);
        System.out.println(bt[x].toString());
        System.out.println("  False negatives: " + falseNeg);
        System.out.println("  False positives: " + falsePos);
        assertEquals(0, falseNeg);
        assertTrue(falsePos < 2 * expErr[x]);
    }
}
Also used : Path(org.apache.hadoop.fs.Path) KeyValue(org.apache.hadoop.hbase.KeyValue) HColumnDescriptor(org.apache.hadoop.hbase.HColumnDescriptor) HFileContextBuilder(org.apache.hadoop.hbase.io.hfile.HFileContextBuilder) HFileContext(org.apache.hadoop.hbase.io.hfile.HFileContext) TreeSet(java.util.TreeSet) FileSystem(org.apache.hadoop.fs.FileSystem) Scan(org.apache.hadoop.hbase.client.Scan) Test(org.junit.Test)

Example 54 with HFileContext

use of org.apache.hadoop.hbase.io.hfile.HFileContext in project hbase by apache.

the class TestStoreFileScannerWithTagCompression method testReseek.

@Test
public void testReseek() throws Exception {
    // write the file
    Path f = new Path(ROOT_DIR, "testReseek");
    HFileContext meta = new HFileContextBuilder().withBlockSize(8 * 1024).withIncludesTags(true).withCompressTags(true).withDataBlockEncoding(DataBlockEncoding.PREFIX).build();
    // Make a store file and write data to it.
    StoreFileWriter writer = new StoreFileWriter.Builder(conf, cacheConf, fs).withFilePath(f).withFileContext(meta).build();
    writeStoreFile(writer);
    writer.close();
    StoreFileReader reader = new StoreFileReader(fs, f, cacheConf, conf);
    StoreFileScanner s = reader.getStoreFileScanner(false, false, false, 0, 0, false);
    try {
        // Now do reseek with empty KV to position to the beginning of the file
        KeyValue k = KeyValueUtil.createFirstOnRow(Bytes.toBytes("k2"));
        s.reseek(k);
        Cell kv = s.next();
        kv = s.next();
        kv = s.next();
        byte[] key5 = Bytes.toBytes("k5");
        assertTrue(Bytes.equals(key5, 0, key5.length, kv.getRowArray(), kv.getRowOffset(), kv.getRowLength()));
        List<Tag> tags = KeyValueUtil.ensureKeyValue(kv).getTags();
        assertEquals(1, tags.size());
        assertEquals("tag3", Bytes.toString(TagUtil.cloneValue(tags.get(0))));
    } finally {
        s.close();
    }
}
Also used : Path(org.apache.hadoop.fs.Path) KeyValue(org.apache.hadoop.hbase.KeyValue) HFileContextBuilder(org.apache.hadoop.hbase.io.hfile.HFileContextBuilder) ArrayBackedTag(org.apache.hadoop.hbase.ArrayBackedTag) Tag(org.apache.hadoop.hbase.Tag) Cell(org.apache.hadoop.hbase.Cell) HFileContext(org.apache.hadoop.hbase.io.hfile.HFileContext) Test(org.junit.Test)

Example 55 with HFileContext

use of org.apache.hadoop.hbase.io.hfile.HFileContext in project phoenix by apache.

the class MultiHfileOutputFormat method createRecordWriter.

/**
     * 
     * @param context
     * @return
     * @throws IOException 
     */
static <V extends Cell> RecordWriter<TableRowkeyPair, V> createRecordWriter(final TaskAttemptContext context) throws IOException {
    // Get the path of the temporary output file
    final Path outputPath = FileOutputFormat.getOutputPath(context);
    final Path outputdir = new FileOutputCommitter(outputPath, context).getWorkPath();
    final Configuration conf = context.getConfiguration();
    final FileSystem fs = outputdir.getFileSystem(conf);
    final long maxsize = conf.getLong(HConstants.HREGION_MAX_FILESIZE, HConstants.DEFAULT_MAX_FILE_SIZE);
    // Invented config.  Add to hbase-*.xml if other than default compression.
    final String defaultCompressionStr = conf.get("hfile.compression", Compression.Algorithm.NONE.getName());
    final Algorithm defaultCompression = AbstractHFileWriter.compressionByName(defaultCompressionStr);
    final boolean compactionExclude = conf.getBoolean("hbase.mapreduce.hfileoutputformat.compaction.exclude", false);
    return new RecordWriter<TableRowkeyPair, V>() {

        // Map of families to writers and how much has been output on the writer.
        private final Map<byte[], WriterLength> writers = new TreeMap<byte[], WriterLength>(Bytes.BYTES_COMPARATOR);

        private byte[] previousRow = HConstants.EMPTY_BYTE_ARRAY;

        private final byte[] now = Bytes.toBytes(System.currentTimeMillis());

        private boolean rollRequested = false;

        @Override
        public void write(TableRowkeyPair row, V cell) throws IOException {
            KeyValue kv = KeyValueUtil.ensureKeyValue(cell);
            // null input == user explicitly wants to flush
            if (row == null && kv == null) {
                rollWriters();
                return;
            }
            // phoenix-2216: start : extract table name from the rowkey
            String tableName = row.getTableName();
            byte[] rowKey = row.getRowkey().get();
            long length = kv.getLength();
            byte[] family = CellUtil.cloneFamily(kv);
            byte[] tableAndFamily = join(tableName, Bytes.toString(family));
            WriterLength wl = this.writers.get(tableAndFamily);
            // If this is a new column family, verify that the directory exists
            if (wl == null) {
                // phoenix-2216: start : create a directory for table and family within the output dir 
                Path tableOutputPath = CsvBulkImportUtil.getOutputPath(outputdir, tableName);
                fs.mkdirs(new Path(tableOutputPath, Bytes.toString(family)));
            // phoenix-2216: end
            }
            // maxsize, we need to roll all the writers
            if (wl != null && wl.written + length >= maxsize) {
                this.rollRequested = true;
            }
            // This can only happen once a row is finished though
            if (rollRequested && Bytes.compareTo(this.previousRow, rowKey) != 0) {
                rollWriters();
            }
            // create a new WAL writer, if necessary
            if (wl == null || wl.writer == null) {
                // phoenix-2216: start : passed even the table name
                wl = getNewWriter(tableName, family, conf);
            // phoenix-2216: end
            }
            // we now have the proper WAL writer. full steam ahead
            kv.updateLatestStamp(this.now);
            wl.writer.append(kv);
            wl.written += length;
            // Copy the row so we know when a row transition.
            this.previousRow = rowKey;
        }

        private void rollWriters() throws IOException {
            for (WriterLength wl : this.writers.values()) {
                if (wl.writer != null) {
                    LOG.info("Writer=" + wl.writer.getPath() + ((wl.written == 0) ? "" : ", wrote=" + wl.written));
                    close(wl.writer);
                }
                wl.writer = null;
                wl.written = 0;
            }
            this.rollRequested = false;
        }

        /* Create a new StoreFile.Writer.
           * @param family
           * @return A WriterLength, containing a new StoreFile.Writer.
           * @throws IOException
           */
        @edu.umd.cs.findbugs.annotations.SuppressWarnings(value = "BX_UNBOXING_IMMEDIATELY_REBOXED", justification = "Not important")
        private WriterLength getNewWriter(final String tableName, byte[] family, Configuration conf) throws IOException {
            WriterLength wl = new WriterLength();
            Path tableOutputPath = CsvBulkImportUtil.getOutputPath(outputdir, tableName);
            Path familydir = new Path(tableOutputPath, Bytes.toString(family));
            // phoenix-2216: start : fetching the configuration properties that were set to the table.
            // create a map from column family to the compression algorithm for the table.
            final Map<byte[], Algorithm> compressionMap = createFamilyCompressionMap(conf, tableName);
            final Map<byte[], BloomType> bloomTypeMap = createFamilyBloomTypeMap(conf, tableName);
            final Map<byte[], Integer> blockSizeMap = createFamilyBlockSizeMap(conf, tableName);
            // phoenix-2216: end
            String dataBlockEncodingStr = conf.get(DATABLOCK_ENCODING_OVERRIDE_CONF_KEY);
            final Map<byte[], DataBlockEncoding> datablockEncodingMap = createFamilyDataBlockEncodingMap(conf, tableName);
            final DataBlockEncoding overriddenEncoding;
            if (dataBlockEncodingStr != null) {
                overriddenEncoding = DataBlockEncoding.valueOf(dataBlockEncodingStr);
            } else {
                overriddenEncoding = null;
            }
            Algorithm compression = compressionMap.get(family);
            compression = compression == null ? defaultCompression : compression;
            BloomType bloomType = bloomTypeMap.get(family);
            bloomType = bloomType == null ? BloomType.NONE : bloomType;
            Integer blockSize = blockSizeMap.get(family);
            blockSize = blockSize == null ? HConstants.DEFAULT_BLOCKSIZE : blockSize;
            DataBlockEncoding encoding = overriddenEncoding;
            encoding = encoding == null ? datablockEncodingMap.get(family) : encoding;
            encoding = encoding == null ? DataBlockEncoding.NONE : encoding;
            Configuration tempConf = new Configuration(conf);
            tempConf.setFloat(HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, 0.0f);
            HFileContextBuilder contextBuilder = new HFileContextBuilder().withCompression(compression).withChecksumType(HStore.getChecksumType(conf)).withBytesPerCheckSum(HStore.getBytesPerChecksum(conf)).withBlockSize(blockSize);
            contextBuilder.withDataBlockEncoding(encoding);
            HFileContext hFileContext = contextBuilder.build();
            wl.writer = new StoreFile.WriterBuilder(conf, new CacheConfig(tempConf), fs).withOutputDir(familydir).withBloomType(bloomType).withComparator(KeyValue.COMPARATOR).withFileContext(hFileContext).build();
            // join and put it in the writers map .
            // phoenix-2216: start : holds a map of writers where the 
            //                       key in the map is a join byte array of table name and family.
            byte[] tableAndFamily = join(tableName, Bytes.toString(family));
            this.writers.put(tableAndFamily, wl);
            // phoenix-2216: end
            return wl;
        }

        private void close(final StoreFile.Writer w) throws IOException {
            if (w != null) {
                w.appendFileInfo(StoreFile.BULKLOAD_TIME_KEY, Bytes.toBytes(System.currentTimeMillis()));
                w.appendFileInfo(StoreFile.BULKLOAD_TASK_KEY, Bytes.toBytes(context.getTaskAttemptID().toString()));
                w.appendFileInfo(StoreFile.MAJOR_COMPACTION_KEY, Bytes.toBytes(true));
                w.appendFileInfo(StoreFile.EXCLUDE_FROM_MINOR_COMPACTION_KEY, Bytes.toBytes(compactionExclude));
                w.appendTrackedTimestampsToMetadata();
                w.close();
            }
        }

        @Override
        public void close(TaskAttemptContext c) throws IOException, InterruptedException {
            for (WriterLength wl : this.writers.values()) {
                close(wl.writer);
            }
        }
    };
}
Also used : DataBlockEncoding(org.apache.hadoop.hbase.io.encoding.DataBlockEncoding) KeyValue(org.apache.hadoop.hbase.KeyValue) Configuration(org.apache.hadoop.conf.Configuration) HFileContextBuilder(org.apache.hadoop.hbase.io.hfile.HFileContextBuilder) RecordWriter(org.apache.hadoop.mapreduce.RecordWriter) TableRowkeyPair(org.apache.phoenix.mapreduce.bulkload.TableRowkeyPair) FileSystem(org.apache.hadoop.fs.FileSystem) CacheConfig(org.apache.hadoop.hbase.io.hfile.CacheConfig) Path(org.apache.hadoop.fs.Path) FileOutputCommitter(org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter) TaskAttemptContext(org.apache.hadoop.mapreduce.TaskAttemptContext) Algorithm(org.apache.hadoop.hbase.io.compress.Compression.Algorithm) HFileContext(org.apache.hadoop.hbase.io.hfile.HFileContext) BloomType(org.apache.hadoop.hbase.regionserver.BloomType) Map(java.util.Map) TreeMap(java.util.TreeMap) RecordWriter(org.apache.hadoop.mapreduce.RecordWriter) AbstractHFileWriter(org.apache.hadoop.hbase.io.hfile.AbstractHFileWriter)

Aggregations

HFileContext (org.apache.hadoop.hbase.io.hfile.HFileContext)55 HFileContextBuilder (org.apache.hadoop.hbase.io.hfile.HFileContextBuilder)51 Path (org.apache.hadoop.fs.Path)34 Test (org.junit.Test)31 KeyValue (org.apache.hadoop.hbase.KeyValue)24 FileSystem (org.apache.hadoop.fs.FileSystem)21 CacheConfig (org.apache.hadoop.hbase.io.hfile.CacheConfig)21 Cell (org.apache.hadoop.hbase.Cell)12 HFile (org.apache.hadoop.hbase.io.hfile.HFile)11 ByteBuffer (java.nio.ByteBuffer)10 Configuration (org.apache.hadoop.conf.Configuration)9 StoreFileWriter (org.apache.hadoop.hbase.regionserver.StoreFileWriter)7 HFileScanner (org.apache.hadoop.hbase.io.hfile.HFileScanner)6 ByteArrayOutputStream (java.io.ByteArrayOutputStream)5 DataOutputStream (java.io.DataOutputStream)5 SingleByteBuff (org.apache.hadoop.hbase.nio.SingleByteBuff)5 HBaseConfiguration (org.apache.hadoop.hbase.HBaseConfiguration)4 HRegionInfo (org.apache.hadoop.hbase.HRegionInfo)4 PrefixTreeCodec (org.apache.hadoop.hbase.codec.prefixtree.PrefixTreeCodec)4 EncodedSeeker (org.apache.hadoop.hbase.io.encoding.DataBlockEncoder.EncodedSeeker)4