Search in sources :

Example 6 with StreamingConnection

use of org.apache.hive.hcatalog.streaming.StreamingConnection in project hive by apache.

the class TestCompactor method testStatsAfterCompactionPartTbl.

/**
   * After each major compaction, stats need to be updated on each column of the
   * table/partition which previously had stats.
   * 1. create a bucketed ORC backed table (Orc is currently required by ACID)
   * 2. populate 2 partitions with data
   * 3. compute stats
   * 4. insert some data into the table using StreamingAPI
   * 5. Trigger major compaction (which should update stats)
   * 6. check that stats have been updated
   * @throws Exception
   * todo:
   * 2. add non-partitioned test
   * 4. add a test with sorted table?
   */
@Test
public void testStatsAfterCompactionPartTbl() throws Exception {
    //as of (8/27/2014) Hive 0.14, ACID/Orc requires HiveInputFormat
    String tblName = "compaction_test";
    String tblNameStg = tblName + "_stg";
    List<String> colNames = Arrays.asList("a", "b");
    executeStatementOnDriver("drop table if exists " + tblName, driver);
    executeStatementOnDriver("drop table if exists " + tblNameStg, driver);
    executeStatementOnDriver("CREATE TABLE " + tblName + "(a INT, b STRING) " + " PARTITIONED BY(bkt INT)" + //currently ACID requires table to be bucketed
    " CLUSTERED BY(a) INTO 4 BUCKETS" + " STORED AS ORC  TBLPROPERTIES ('transactional'='true')", driver);
    executeStatementOnDriver("CREATE EXTERNAL TABLE " + tblNameStg + "(a INT, b STRING)" + " ROW FORMAT DELIMITED FIELDS TERMINATED BY '\\t' LINES TERMINATED BY '\\n'" + " STORED AS TEXTFILE" + " LOCATION '" + stagingFolder.newFolder().toURI().getPath() + "'", driver);
    executeStatementOnDriver("load data local inpath '" + BASIC_FILE_NAME + "' overwrite into table " + tblNameStg, driver);
    execSelectAndDumpData("select * from " + tblNameStg, driver, "Dumping data for " + tblNameStg + " after load:");
    executeStatementOnDriver("FROM " + tblNameStg + " INSERT INTO TABLE " + tblName + " PARTITION(bkt=0) " + "SELECT a, b where a < 2", driver);
    executeStatementOnDriver("FROM " + tblNameStg + " INSERT INTO TABLE " + tblName + " PARTITION(bkt=1) " + "SELECT a, b where a >= 2", driver);
    execSelectAndDumpData("select * from " + tblName, driver, "Dumping data for " + tblName + " after load:");
    TxnStore txnHandler = TxnUtils.getTxnStore(conf);
    CompactionInfo ci = new CompactionInfo("default", tblName, "bkt=0", CompactionType.MAJOR);
    LOG.debug("List of stats columns before analyze Part1: " + txnHandler.findColumnsWithStats(ci));
    Worker.StatsUpdater su = Worker.StatsUpdater.init(ci, colNames, conf, System.getProperty("user.name"));
    //compute stats before compaction
    su.gatherStats();
    LOG.debug("List of stats columns after analyze Part1: " + txnHandler.findColumnsWithStats(ci));
    CompactionInfo ciPart2 = new CompactionInfo("default", tblName, "bkt=1", CompactionType.MAJOR);
    LOG.debug("List of stats columns before analyze Part2: " + txnHandler.findColumnsWithStats(ci));
    su = Worker.StatsUpdater.init(ciPart2, colNames, conf, System.getProperty("user.name"));
    //compute stats before compaction
    su.gatherStats();
    LOG.debug("List of stats columns after analyze Part2: " + txnHandler.findColumnsWithStats(ci));
    //now make sure we get the stats we expect for partition we are going to add data to later
    Map<String, List<ColumnStatisticsObj>> stats = msClient.getPartitionColumnStatistics(ci.dbname, ci.tableName, Arrays.asList(ci.partName), colNames);
    List<ColumnStatisticsObj> colStats = stats.get(ci.partName);
    Assert.assertNotNull("No stats found for partition " + ci.partName, colStats);
    Assert.assertEquals("Expected column 'a' at index 0", "a", colStats.get(0).getColName());
    Assert.assertEquals("Expected column 'b' at index 1", "b", colStats.get(1).getColName());
    LongColumnStatsData colAStats = colStats.get(0).getStatsData().getLongStats();
    Assert.assertEquals("lowValue a", 1, colAStats.getLowValue());
    Assert.assertEquals("highValue a", 1, colAStats.getHighValue());
    Assert.assertEquals("numNulls a", 0, colAStats.getNumNulls());
    Assert.assertEquals("numNdv a", 1, colAStats.getNumDVs());
    StringColumnStatsData colBStats = colStats.get(1).getStatsData().getStringStats();
    Assert.assertEquals("maxColLen b", 3, colBStats.getMaxColLen());
    Assert.assertEquals("avgColLen b", 3.0, colBStats.getAvgColLen(), 0.01);
    Assert.assertEquals("numNulls b", 0, colBStats.getNumNulls());
    Assert.assertEquals("nunDVs", 2, colBStats.getNumDVs());
    //now save stats for partition we won't modify
    stats = msClient.getPartitionColumnStatistics(ciPart2.dbname, ciPart2.tableName, Arrays.asList(ciPart2.partName), colNames);
    colStats = stats.get(ciPart2.partName);
    LongColumnStatsData colAStatsPart2 = colStats.get(0).getStatsData().getLongStats();
    StringColumnStatsData colBStatsPart2 = colStats.get(1).getStatsData().getStringStats();
    HiveEndPoint endPt = new HiveEndPoint(null, ci.dbname, ci.tableName, Arrays.asList("0"));
    DelimitedInputWriter writer = new DelimitedInputWriter(new String[] { "a", "b" }, ",", endPt);
    /*next call will eventually end up in HiveEndPoint.createPartitionIfNotExists() which
    makes an operation on Driver
    * and starts it's own CliSessionState and then closes it, which removes it from ThreadLoacal;
    * thus the session
    * created in this class is gone after this; I fixed it in HiveEndPoint*/
    StreamingConnection connection = endPt.newConnection(true, "UT_" + Thread.currentThread().getName());
    TransactionBatch txnBatch = connection.fetchTransactionBatch(2, writer);
    txnBatch.beginNextTransaction();
    Assert.assertEquals(TransactionBatch.TxnState.OPEN, txnBatch.getCurrentTransactionState());
    txnBatch.write("50,Kiev".getBytes());
    txnBatch.write("51,St. Petersburg".getBytes());
    txnBatch.write("44,Boston".getBytes());
    txnBatch.commit();
    txnBatch.beginNextTransaction();
    txnBatch.write("52,Tel Aviv".getBytes());
    txnBatch.write("53,Atlantis".getBytes());
    txnBatch.write("53,Boston".getBytes());
    txnBatch.commit();
    txnBatch.close();
    connection.close();
    execSelectAndDumpData("select * from " + ci.getFullTableName(), driver, ci.getFullTableName());
    //so now we have written some new data to bkt=0 and it shows up
    CompactionRequest rqst = new CompactionRequest(ci.dbname, ci.tableName, CompactionType.MAJOR);
    rqst.setPartitionname(ci.partName);
    txnHandler.compact(rqst);
    Worker t = new Worker();
    t.setThreadId((int) t.getId());
    t.setHiveConf(conf);
    AtomicBoolean stop = new AtomicBoolean();
    AtomicBoolean looped = new AtomicBoolean();
    stop.set(true);
    t.init(stop, looped);
    t.run();
    ShowCompactResponse rsp = txnHandler.showCompact(new ShowCompactRequest());
    List<ShowCompactResponseElement> compacts = rsp.getCompacts();
    if (1 != compacts.size()) {
        Assert.fail("Expecting 1 file and found " + compacts.size() + " files " + compacts.toString());
    }
    Assert.assertEquals("ready for cleaning", compacts.get(0).getState());
    stats = msClient.getPartitionColumnStatistics(ci.dbname, ci.tableName, Arrays.asList(ci.partName), colNames);
    colStats = stats.get(ci.partName);
    Assert.assertNotNull("No stats found for partition " + ci.partName, colStats);
    Assert.assertEquals("Expected column 'a' at index 0", "a", colStats.get(0).getColName());
    Assert.assertEquals("Expected column 'b' at index 1", "b", colStats.get(1).getColName());
    colAStats = colStats.get(0).getStatsData().getLongStats();
    Assert.assertEquals("lowValue a", 1, colAStats.getLowValue());
    Assert.assertEquals("highValue a", 53, colAStats.getHighValue());
    Assert.assertEquals("numNulls a", 0, colAStats.getNumNulls());
    Assert.assertEquals("numNdv a", 6, colAStats.getNumDVs());
    colBStats = colStats.get(1).getStatsData().getStringStats();
    Assert.assertEquals("maxColLen b", 14, colBStats.getMaxColLen());
    //cast it to long to get rid of periodic decimal
    Assert.assertEquals("avgColLen b", (long) 6.1111111111, (long) colBStats.getAvgColLen());
    Assert.assertEquals("numNulls b", 0, colBStats.getNumNulls());
    Assert.assertEquals("nunDVs", 10, colBStats.getNumDVs());
    //now check that stats for partition we didn't modify did not change
    stats = msClient.getPartitionColumnStatistics(ciPart2.dbname, ciPart2.tableName, Arrays.asList(ciPart2.partName), colNames);
    colStats = stats.get(ciPart2.partName);
    Assert.assertEquals("Expected stats for " + ciPart2.partName + " to stay the same", colAStatsPart2, colStats.get(0).getStatsData().getLongStats());
    Assert.assertEquals("Expected stats for " + ciPart2.partName + " to stay the same", colBStatsPart2, colStats.get(1).getStatsData().getStringStats());
}
Also used : TransactionBatch(org.apache.hive.hcatalog.streaming.TransactionBatch) StreamingConnection(org.apache.hive.hcatalog.streaming.StreamingConnection) StringColumnStatsData(org.apache.hadoop.hive.metastore.api.StringColumnStatsData) LongColumnStatsData(org.apache.hadoop.hive.metastore.api.LongColumnStatsData) ColumnStatisticsObj(org.apache.hadoop.hive.metastore.api.ColumnStatisticsObj) AtomicBoolean(java.util.concurrent.atomic.AtomicBoolean) ShowCompactResponse(org.apache.hadoop.hive.metastore.api.ShowCompactResponse) DelimitedInputWriter(org.apache.hive.hcatalog.streaming.DelimitedInputWriter) HiveEndPoint(org.apache.hive.hcatalog.streaming.HiveEndPoint) CompactionInfo(org.apache.hadoop.hive.metastore.txn.CompactionInfo) List(java.util.List) ArrayList(java.util.ArrayList) ValidTxnList(org.apache.hadoop.hive.common.ValidTxnList) ShowCompactRequest(org.apache.hadoop.hive.metastore.api.ShowCompactRequest) TxnStore(org.apache.hadoop.hive.metastore.txn.TxnStore) CompactionRequest(org.apache.hadoop.hive.metastore.api.CompactionRequest) ShowCompactResponseElement(org.apache.hadoop.hive.metastore.api.ShowCompactResponseElement) Test(org.junit.Test)

Example 7 with StreamingConnection

use of org.apache.hive.hcatalog.streaming.StreamingConnection in project hive by apache.

the class TestCompactor method minorCompactAfterAbort.

@Test
public void minorCompactAfterAbort() throws Exception {
    String agentInfo = "UT_" + Thread.currentThread().getName();
    String dbName = "default";
    String tblName = "cws";
    List<String> colNames = Arrays.asList("a", "b");
    String columnNamesProperty = "a,b";
    String columnTypesProperty = "int:string";
    executeStatementOnDriver("drop table if exists " + tblName, driver);
    executeStatementOnDriver("CREATE TABLE " + tblName + "(a INT, b STRING) " + //currently ACID requires table to be bucketed
    " CLUSTERED BY(a) INTO 1 BUCKETS" + " STORED AS ORC  TBLPROPERTIES ('transactional'='true')", driver);
    HiveEndPoint endPt = new HiveEndPoint(null, dbName, tblName, null);
    DelimitedInputWriter writer = new DelimitedInputWriter(new String[] { "a", "b" }, ",", endPt);
    StreamingConnection connection = endPt.newConnection(false, "UT_" + Thread.currentThread().getName());
    try {
        // Write a couple of batches
        for (int i = 0; i < 2; i++) {
            writeBatch(connection, writer, false);
        }
        // Start a third batch, abort everything, don't properly close it
        TransactionBatch txnBatch = connection.fetchTransactionBatch(2, writer);
        txnBatch.beginNextTransaction();
        txnBatch.abort();
        txnBatch.beginNextTransaction();
        txnBatch.abort();
        // Now, compact
        TxnStore txnHandler = TxnUtils.getTxnStore(conf);
        txnHandler.compact(new CompactionRequest(dbName, tblName, CompactionType.MINOR));
        Worker t = new Worker();
        t.setThreadId((int) t.getId());
        t.setHiveConf(conf);
        AtomicBoolean stop = new AtomicBoolean(true);
        AtomicBoolean looped = new AtomicBoolean();
        t.init(stop, looped);
        t.run();
        // Find the location of the table
        IMetaStoreClient msClient = new HiveMetaStoreClient(conf);
        Table table = msClient.getTable(dbName, tblName);
        FileSystem fs = FileSystem.get(conf);
        FileStatus[] stat = fs.listStatus(new Path(table.getSd().getLocation()), AcidUtils.deltaFileFilter);
        String[] names = new String[stat.length];
        Path resultDelta = null;
        for (int i = 0; i < names.length; i++) {
            names[i] = stat[i].getPath().getName();
            if (names[i].equals("delta_0000001_0000004")) {
                resultDelta = stat[i].getPath();
            }
        }
        Arrays.sort(names);
        String[] expected = new String[] { "delta_0000001_0000002", "delta_0000001_0000004", "delta_0000003_0000004" };
        if (!Arrays.deepEquals(expected, names)) {
            Assert.fail("Expected: " + Arrays.toString(expected) + ", found: " + Arrays.toString(names));
        }
        checkExpectedTxnsPresent(null, new Path[] { resultDelta }, columnNamesProperty, columnTypesProperty, 0, 1L, 4L);
    } finally {
        connection.close();
    }
}
Also used : Path(org.apache.hadoop.fs.Path) HiveMetaStoreClient(org.apache.hadoop.hive.metastore.HiveMetaStoreClient) Table(org.apache.hadoop.hive.metastore.api.Table) FileStatus(org.apache.hadoop.fs.FileStatus) TransactionBatch(org.apache.hive.hcatalog.streaming.TransactionBatch) StreamingConnection(org.apache.hive.hcatalog.streaming.StreamingConnection) IMetaStoreClient(org.apache.hadoop.hive.metastore.IMetaStoreClient) HiveEndPoint(org.apache.hive.hcatalog.streaming.HiveEndPoint) AtomicBoolean(java.util.concurrent.atomic.AtomicBoolean) FileSystem(org.apache.hadoop.fs.FileSystem) DelimitedInputWriter(org.apache.hive.hcatalog.streaming.DelimitedInputWriter) HiveEndPoint(org.apache.hive.hcatalog.streaming.HiveEndPoint) TxnStore(org.apache.hadoop.hive.metastore.txn.TxnStore) CompactionRequest(org.apache.hadoop.hive.metastore.api.CompactionRequest) Test(org.junit.Test)

Aggregations

AtomicBoolean (java.util.concurrent.atomic.AtomicBoolean)7 CompactionRequest (org.apache.hadoop.hive.metastore.api.CompactionRequest)7 TxnStore (org.apache.hadoop.hive.metastore.txn.TxnStore)7 DelimitedInputWriter (org.apache.hive.hcatalog.streaming.DelimitedInputWriter)7 HiveEndPoint (org.apache.hive.hcatalog.streaming.HiveEndPoint)7 StreamingConnection (org.apache.hive.hcatalog.streaming.StreamingConnection)7 Test (org.junit.Test)7 FileStatus (org.apache.hadoop.fs.FileStatus)6 FileSystem (org.apache.hadoop.fs.FileSystem)6 Path (org.apache.hadoop.fs.Path)6 HiveMetaStoreClient (org.apache.hadoop.hive.metastore.HiveMetaStoreClient)6 IMetaStoreClient (org.apache.hadoop.hive.metastore.IMetaStoreClient)6 Table (org.apache.hadoop.hive.metastore.api.Table)6 TransactionBatch (org.apache.hive.hcatalog.streaming.TransactionBatch)3 ArrayList (java.util.ArrayList)1 List (java.util.List)1 ValidTxnList (org.apache.hadoop.hive.common.ValidTxnList)1 ColumnStatisticsObj (org.apache.hadoop.hive.metastore.api.ColumnStatisticsObj)1 LongColumnStatsData (org.apache.hadoop.hive.metastore.api.LongColumnStatsData)1 ShowCompactRequest (org.apache.hadoop.hive.metastore.api.ShowCompactRequest)1