I'm trying to create HFiles programmatically and loading them in a running HBase instance. I found a lot of info in HFileOutputFormat and in LoadIncrementalHFiles
I managed to create the new HFile, send it to the cluster. In the cluster web interface the new store file appears but the new keyrange is unavailable.
InputStream stream = ProgrammaticHFileGeneration.class.getResourceAsStream("ga-hourly.txt");
BufferedReader reader = new BufferedReader(new InputStreamReader(stream));
String line = null;
Map<byte[], String> rowValues = new HashMap<byte[], String>();
while((line = reader.readLine())!=null) {
String[] vals = line.split(",");
String row = new StringBuilder(vals[0]).append(".").append(vals[1]).append(".").append(vals[2]).append(".").append(vals[3]).toString();
rowValues.put(row.getBytes(), line);
}
List<byte[]> keys = new ArrayList<byte[]>(rowValues.keySet());
Collections.sort(keys, byteArrComparator);
HBaseTestingUtility testingUtility = new HBaseTestingUtility();
testingUtility.startMiniCluster();
testingUtility.createTable("table".getBytes(), "data".getBytes());
Writer writer = new HFile.Writer(testingUtility.getTestFileSystem(),
new Path("/tmp/hfiles/data/hfile"),
HFile.DEFAULT_BLOCKSIZE, Compression.Algorithm.NONE, KeyValue.KEY_COMPARATOR);
for(byte[] key:keys) {
writer.append(new KeyValue(key, "data".getBytes(), "d".getBytes(), rowValues.get(key).getBytes()));
}
writer.appendFileInfo(StoreFile.BULKLOAD_TIME_KEY, Bytes.toBytes(System.currentTimeMillis()));
writer.appendFileInfo(StoreFile.MAJOR_COMPACTION_KEY, Bytes.toBytes(true));
writer.close();
Configuration conf = testingUtility.getConfiguration();
LoadIncrementalHFiles loadTool = new LoadIncrementalHFiles(conf);
HTable hTable = new HTable(conf, "table".getBytes());
loadTool.doBulkLoad(new Path("/tmp/hfiles"), hTable);
ResultScanner scanner = hTable.getScanner("data".getBytes());
Result next = null;
System.out.println("Scanning");
while((next = scanner.next()) != null) {
System.out.format("%s %s\n", new String(next.getRow()), new String(next.getValue("data".getBytes(), "d".getBytes())));
}
Did anyone actually make this work ? I have a compilable / testable version up on my github
