@BeforeClass
public static void setUpBeforeClass() throws Exception {
// Start up our mini cluster on top of an 0.92 root.dir that has data from
// a 0.92 hbase run -- it has a table with 100 rows in it -- and see if
// we can migrate from 0.92
TEST_UTIL.startMiniZKCluster();
TEST_UTIL.startMiniDFSCluster(1);
Path testdir = TEST_UTIL.getDataTestDir("TestMetaMigrationConvertToPB");
// Untar our test dir.
File untar = untar(new File(testdir.toString()));
// Now copy the untar up into hdfs so when we start hbase, we'll run from it.
Configuration conf = TEST_UTIL.getConfiguration();
FsShell shell = new FsShell(conf);
FileSystem fs = FileSystem.get(conf);
// find where hbase will root itself, so we can copy filesystem there
Path hbaseRootDir = TEST_UTIL.getDefaultRootDirPath();
if (!fs.isDirectory(hbaseRootDir.getParent())) {
// mkdir at first
fs.mkdirs(hbaseRootDir.getParent());
}
doFsCommand(shell, new String[] {"-put", untar.toURI().toString(), hbaseRootDir.toString()});
// windows fix: tgz file has .META. directory renamed as -META- since the original is an illegal
// name under windows. So we rename it back. See
// src/test/data//TestMetaMigrationConvertingToPB.README and
// https://issues.apache.org/jira/browse/HBASE-6821
doFsCommand(
shell,
new String[] {
"-mv",
new Path(hbaseRootDir, "-META-").toString(),
new Path(hbaseRootDir, ".META.").toString()
});
// See whats in minihdfs.
doFsCommand(shell, new String[] {"-lsr", "/"});
TEST_UTIL.startMiniHBaseCluster(1, 1);
// Assert we are running against the copied-up filesystem. The copied-up
// rootdir should have had a table named 'TestTable' in it. Assert it
// present.
HTable t = new HTable(TEST_UTIL.getConfiguration(), TESTTABLE);
ResultScanner scanner = t.getScanner(new Scan());
int count = 0;
while (scanner.next() != null) {
count++;
}
// Assert that we find all 100 rows that are in the data we loaded. If
// so then we must have migrated it from 0.90 to 0.92.
Assert.assertEquals(ROW_COUNT, count);
scanner.close();
t.close();
}
@Before
public void setUp() throws Exception {
File tempDir = Files.createTempDir();
tempDir.deleteOnExit();
htu = HBaseTestingUtility.createLocalHTU();
try {
htu.cleanupTestDir();
htu.startMiniZKCluster();
htu.startMiniHBaseCluster(1, 1);
try {
htu.deleteTable(Bytes.toBytes(tableName));
} catch (Exception e) {
Log.info(" - no table " + tableName + " found");
}
htu.createTable(Bytes.toBytes(tableName), colFam);
dao =
new Hbase1OffsetStore.Builder()
.setHbaseConfiguration(htu.getConfiguration())
.setOffsetTable(tableName)
.build();
} catch (Exception e1) {
throw new RuntimeException(e1);
}
KOM =
new KafkaOffsetManager.Builder()
.setOffsetManager(dao)
.setKafkaBrokerList("localhost:" + kafkaRule.kafkaBrokerPort())
.setGroupID(testGroupID)
.setTopic(testTopicName)
.build();
}
@Test
public void testReducerNumEstimation() throws Exception {
// Skip the test for Tez. Tez use a different mechanism.
// Equivalent test is in TestTezAutoParallelism
Assume.assumeTrue("Skip this test for TEZ", Util.isMapredExecType(cluster.getExecType()));
// use the estimation
Configuration conf = HBaseConfiguration.create(new Configuration());
HBaseTestingUtility util = new HBaseTestingUtility(conf);
int clientPort = util.startMiniZKCluster().getClientPort();
util.startMiniHBaseCluster(1, 1);
String query = "a = load '/passwd';" + "b = group a by $0;" + "store b into 'output';";
PigServer ps = new PigServer(cluster.getExecType(), cluster.getProperties());
PhysicalPlan pp = Util.buildPp(ps, query);
MROperPlan mrPlan = Util.buildMRPlan(pp, pc);
pc.getConf().setProperty("pig.exec.reducers.bytes.per.reducer", "100");
pc.getConf().setProperty("pig.exec.reducers.max", "10");
pc.getConf().setProperty(HConstants.ZOOKEEPER_CLIENT_PORT, Integer.toString(clientPort));
ConfigurationValidator.validatePigProperties(pc.getProperties());
conf = ConfigurationUtil.toConfiguration(pc.getProperties());
JobControlCompiler jcc = new JobControlCompiler(pc, conf);
JobControl jc = jcc.compile(mrPlan, "Test");
Job job = jc.getWaitingJobs().get(0);
long reducer =
Math.min(
(long) Math.ceil(new File("test/org/apache/pig/test/data/passwd").length() / 100.0),
10);
Util.assertParallelValues(-1, -1, reducer, reducer, job.getJobConf());
// use the PARALLEL key word, it will override the estimated reducer number
query = "a = load '/passwd';" + "b = group a by $0 PARALLEL 2;" + "store b into 'output';";
pp = Util.buildPp(ps, query);
mrPlan = Util.buildMRPlan(pp, pc);
pc.getConf().setProperty("pig.exec.reducers.bytes.per.reducer", "100");
pc.getConf().setProperty("pig.exec.reducers.max", "10");
ConfigurationValidator.validatePigProperties(pc.getProperties());
conf = ConfigurationUtil.toConfiguration(pc.getProperties());
jcc = new JobControlCompiler(pc, conf);
jc = jcc.compile(mrPlan, "Test");
job = jc.getWaitingJobs().get(0);
Util.assertParallelValues(-1, 2, -1, 2, job.getJobConf());
final byte[] COLUMNFAMILY = Bytes.toBytes("pig");
util.createTable(Bytes.toBytesBinary("test_table"), COLUMNFAMILY);
// the estimation won't take effect when it apply to non-dfs or the files doesn't exist, such as
// hbase
query =
"a = load 'hbase://test_table' using org.apache.pig.backend.hadoop.hbase.HBaseStorage('c:f1 c:f2');"
+ "b = group a by $0 ;"
+ "store b into 'output';";
pp = Util.buildPp(ps, query);
mrPlan = Util.buildMRPlan(pp, pc);
pc.getConf().setProperty("pig.exec.reducers.bytes.per.reducer", "100");
pc.getConf().setProperty("pig.exec.reducers.max", "10");
ConfigurationValidator.validatePigProperties(pc.getProperties());
conf = ConfigurationUtil.toConfiguration(pc.getProperties());
jcc = new JobControlCompiler(pc, conf);
jc = jcc.compile(mrPlan, "Test");
job = jc.getWaitingJobs().get(0);
Util.assertParallelValues(-1, -1, 1, 1, job.getJobConf());
util.deleteTable(Bytes.toBytesBinary("test_table"));
// In HBase 0.90.1 and above we can use util.shutdownMiniHBaseCluster()
// here instead.
MiniHBaseCluster hbc = util.getHBaseCluster();
if (hbc != null) {
hbc.shutdown();
hbc.join();
}
util.shutdownMiniZKCluster();
}