Examples with ZooKeeperWatcher org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher used on opensource projects

Example 66 with ZooKeeperWatcher

use of org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher in project hbase by apache.

the class TestDistributedLogSplitting method testLogReplayWithMetaRSDown.

@Ignore("DLR is broken by HBASE-12751")
@Test(timeout = 300000)
public void testLogReplayWithMetaRSDown() throws Exception {
    LOG.info("testRecoveredEditsReplayWithMetaRSDown");
    conf.setBoolean(HConstants.DISTRIBUTED_LOG_REPLAY_KEY, true);
    startCluster(NUM_RS);
    final int NUM_REGIONS_TO_CREATE = 40;
    final int NUM_LOG_LINES = 1000;
    // turn off load balancing to prevent regions from moving around otherwise
    // they will consume recovered.edits
    master.balanceSwitch(false);
    final ZooKeeperWatcher zkw = new ZooKeeperWatcher(conf, "table-creation", null);
    Table ht = installTable(zkw, "table", "family", NUM_REGIONS_TO_CREATE);
    try {
        HRegionServer hrs = findRSToKill(true, "table");
        List<HRegionInfo> regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());
        makeWAL(hrs, regions, "table", "family", NUM_LOG_LINES, 100);
        this.abortRSAndVerifyRecovery(hrs, ht, zkw, NUM_REGIONS_TO_CREATE, NUM_LOG_LINES);
    } finally {
        if (ht != null)
            ht.close();
        if (zkw != null)
            zkw.close();
    }
}

Example 67 with ZooKeeperWatcher

use of org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher in project hbase by apache.

the class TestDistributedLogSplitting method testMarkRegionsRecoveringInZK.

@Ignore("DLR is broken by HBASE-12751")
@Test(timeout = 300000)
public void testMarkRegionsRecoveringInZK() throws Exception {
    LOG.info("testMarkRegionsRecoveringInZK");
    conf.setBoolean(HConstants.DISTRIBUTED_LOG_REPLAY_KEY, true);
    startCluster(NUM_RS);
    master.balanceSwitch(false);
    List<RegionServerThread> rsts = cluster.getLiveRegionServerThreads();
    final ZooKeeperWatcher zkw = master.getZooKeeper();
    Table ht = installTable(zkw, "table", "family", 40);
    try {
        final SplitLogManager slm = master.getMasterWalManager().getSplitLogManager();
        Set<HRegionInfo> regionSet = new HashSet<>();
        HRegionInfo region = null;
        HRegionServer hrs = null;
        ServerName firstFailedServer = null;
        ServerName secondFailedServer = null;
        for (int i = 0; i < NUM_RS; i++) {
            hrs = rsts.get(i).getRegionServer();
            List<HRegionInfo> regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());
            if (regions.isEmpty())
                continue;
            region = regions.get(0);
            regionSet.add(region);
            firstFailedServer = hrs.getServerName();
            secondFailedServer = rsts.get((i + 1) % NUM_RS).getRegionServer().getServerName();
            break;
        }
        slm.markRegionsRecovering(firstFailedServer, regionSet);
        slm.markRegionsRecovering(secondFailedServer, regionSet);
        List<String> recoveringRegions = ZKUtil.listChildrenNoWatch(zkw, ZKUtil.joinZNode(zkw.znodePaths.recoveringRegionsZNode, region.getEncodedName()));
        assertEquals(recoveringRegions.size(), 2);
        // wait for splitLogWorker to mark them up because there is no WAL files recorded in ZK
        final HRegionServer tmphrs = hrs;
        TEST_UTIL.waitFor(60000, 1000, new Waiter.Predicate<Exception>() {

            @Override
            public boolean evaluate() throws Exception {
                return (tmphrs.getRecoveringRegions().isEmpty());
            }
        });
    } finally {
        if (ht != null)
            ht.close();
        if (zkw != null)
            zkw.close();
    }
}

Example 68 with ZooKeeperWatcher

use of org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher in project hbase by apache.

the class TestDistributedLogSplitting method testMetaRecoveryInZK.

@Ignore("DLR is broken by HBASE-12751")
@Test(timeout = 300000)
public void testMetaRecoveryInZK() throws Exception {
    LOG.info("testMetaRecoveryInZK");
    conf.setBoolean(HConstants.DISTRIBUTED_LOG_REPLAY_KEY, true);
    startCluster(NUM_RS);
    // turn off load balancing to prevent regions from moving around otherwise
    // they will consume recovered.edits
    master.balanceSwitch(false);
    final ZooKeeperWatcher zkw = new ZooKeeperWatcher(conf, "table-creation", null);
    // only testing meta recovery in ZK operation
    HRegionServer hrs = findRSToKill(true, null);
    List<HRegionInfo> regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());
    LOG.info("#regions = " + regions.size());
    Set<HRegionInfo> tmpRegions = new HashSet<>();
    tmpRegions.add(HRegionInfo.FIRST_META_REGIONINFO);
    master.getMasterWalManager().prepareLogReplay(hrs.getServerName(), tmpRegions);
    Set<HRegionInfo> userRegionSet = new HashSet<>();
    userRegionSet.addAll(regions);
    master.getMasterWalManager().prepareLogReplay(hrs.getServerName(), userRegionSet);
    boolean isMetaRegionInRecovery = false;
    List<String> recoveringRegions = zkw.getRecoverableZooKeeper().getChildren(zkw.znodePaths.recoveringRegionsZNode, false);
    for (String curEncodedRegionName : recoveringRegions) {
        if (curEncodedRegionName.equals(HRegionInfo.FIRST_META_REGIONINFO.getEncodedName())) {
            isMetaRegionInRecovery = true;
            break;
        }
    }
    assertTrue(isMetaRegionInRecovery);
    master.getMasterWalManager().splitMetaLog(hrs.getServerName());
    isMetaRegionInRecovery = false;
    recoveringRegions = zkw.getRecoverableZooKeeper().getChildren(zkw.znodePaths.recoveringRegionsZNode, false);
    for (String curEncodedRegionName : recoveringRegions) {
        if (curEncodedRegionName.equals(HRegionInfo.FIRST_META_REGIONINFO.getEncodedName())) {
            isMetaRegionInRecovery = true;
            break;
        }
    }
    // meta region should be recovered
    assertFalse(isMetaRegionInRecovery);
    zkw.close();
}

Example 69 with ZooKeeperWatcher

use of org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher in project hbase by apache.

the class TestDistributedLogSplitting method testMasterStartsUpWithLogSplittingWork.

@Ignore("DLR is broken by HBASE-12751")
@Test(timeout = 300000)
public void testMasterStartsUpWithLogSplittingWork() throws Exception {
    LOG.info("testMasterStartsUpWithLogSplittingWork");
    conf.setBoolean(HConstants.DISTRIBUTED_LOG_REPLAY_KEY, false);
    conf.setInt(ServerManager.WAIT_ON_REGIONSERVERS_MINTOSTART, NUM_RS - 1);
    startCluster(NUM_RS);
    final int NUM_REGIONS_TO_CREATE = 40;
    final int NUM_LOG_LINES = 1000;
    // turn off load balancing to prevent regions from moving around otherwise
    // they will consume recovered.edits
    master.balanceSwitch(false);
    final ZooKeeperWatcher zkw = new ZooKeeperWatcher(conf, "table-creation", null);
    Table ht = installTable(zkw, "table", "family", NUM_REGIONS_TO_CREATE);
    try {
        HRegionServer hrs = findRSToKill(false, "table");
        List<HRegionInfo> regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());
        makeWAL(hrs, regions, "table", "family", NUM_LOG_LINES, 100);
        // abort master
        abortMaster(cluster);
        // abort RS
        LOG.info("Aborting region server: " + hrs.getServerName());
        hrs.abort("testing");
        // wait for abort completes
        TEST_UTIL.waitFor(120000, 200, new Waiter.Predicate<Exception>() {

            @Override
            public boolean evaluate() throws Exception {
                return (cluster.getLiveRegionServerThreads().size() <= (NUM_RS - 1));
            }
        });
        Thread.sleep(2000);
        LOG.info("Current Open Regions:" + HBaseTestingUtility.getAllOnlineRegions(cluster).size());
        // wait for abort completes
        TEST_UTIL.waitFor(120000, 200, new Waiter.Predicate<Exception>() {

            @Override
            public boolean evaluate() throws Exception {
                return (HBaseTestingUtility.getAllOnlineRegions(cluster).size() >= (NUM_REGIONS_TO_CREATE + 1));
            }
        });
        LOG.info("Current Open Regions After Master Node Starts Up:" + HBaseTestingUtility.getAllOnlineRegions(cluster).size());
        assertEquals(NUM_LOG_LINES, TEST_UTIL.countRows(ht));
    } finally {
        if (ht != null)
            ht.close();
        if (zkw != null)
            zkw.close();
    }
}

Example 70 with ZooKeeperWatcher

use of org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher in project hbase by apache.

the class TestDistributedLogSplitting method testSameVersionUpdatesRecovery.

@Ignore("DLR is broken by HBASE-12751")
@Test(timeout = 300000)
public void testSameVersionUpdatesRecovery() throws Exception {
    LOG.info("testSameVersionUpdatesRecovery");
    conf.setLong("hbase.regionserver.hlog.blocksize", 15 * 1024);
    conf.setBoolean(HConstants.DISTRIBUTED_LOG_REPLAY_KEY, true);
    startCluster(NUM_RS);
    final AtomicLong sequenceId = new AtomicLong(100);
    final int NUM_REGIONS_TO_CREATE = 40;
    final int NUM_LOG_LINES = 1000;
    // turn off load balancing to prevent regions from moving around otherwise
    // they will consume recovered.edits
    master.balanceSwitch(false);
    List<RegionServerThread> rsts = cluster.getLiveRegionServerThreads();
    final ZooKeeperWatcher zkw = new ZooKeeperWatcher(conf, "table-creation", null);
    Table ht = installTable(zkw, name.getMethodName(), "family", NUM_REGIONS_TO_CREATE);
    try {
        List<HRegionInfo> regions = null;
        HRegionServer hrs = null;
        for (int i = 0; i < NUM_RS; i++) {
            boolean isCarryingMeta = false;
            hrs = rsts.get(i).getRegionServer();
            regions = ProtobufUtil.getOnlineRegions(hrs.getRSRpcServices());
            for (HRegionInfo region : regions) {
                if (region.isMetaRegion()) {
                    isCarryingMeta = true;
                    break;
                }
            }
            if (isCarryingMeta) {
                continue;
            }
            break;
        }
        LOG.info("#regions = " + regions.size());
        Iterator<HRegionInfo> it = regions.iterator();
        while (it.hasNext()) {
            HRegionInfo region = it.next();
            if (region.isMetaTable() || region.getEncodedName().equals(HRegionInfo.FIRST_META_REGIONINFO.getEncodedName())) {
                it.remove();
            }
        }
        if (regions.isEmpty())
            return;
        HRegionInfo curRegionInfo = regions.get(0);
        byte[] startRow = curRegionInfo.getStartKey();
        if (startRow == null || startRow.length == 0) {
            startRow = new byte[] { 0, 0, 0, 0, 1 };
        }
        byte[] row = Bytes.incrementBytes(startRow, 1);
        // use last 5 bytes because HBaseTestingUtility.createMultiRegions use 5 bytes key
        row = Arrays.copyOfRange(row, 3, 8);
        long value = 0;
        TableName tableName = TableName.valueOf(name.getMethodName());
        byte[] family = Bytes.toBytes("family");
        byte[] qualifier = Bytes.toBytes("c1");
        long timeStamp = System.currentTimeMillis();
        HTableDescriptor htd = new HTableDescriptor(tableName);
        htd.addFamily(new HColumnDescriptor(family));
        final WAL wal = hrs.getWAL(curRegionInfo);
        for (int i = 0; i < NUM_LOG_LINES; i += 1) {
            WALEdit e = new WALEdit();
            value++;
            e.add(new KeyValue(row, family, qualifier, timeStamp, Bytes.toBytes(value)));
            wal.append(curRegionInfo, new WALKey(curRegionInfo.getEncodedNameAsBytes(), tableName, System.currentTimeMillis()), e, true);
        }
        wal.sync();
        wal.shutdown();
        // wait for abort completes
        this.abortRSAndWaitForRecovery(hrs, zkw, NUM_REGIONS_TO_CREATE);
        // verify we got the last value
        LOG.info("Verification Starts...");
        Get g = new Get(row);
        Result r = ht.get(g);
        long theStoredVal = Bytes.toLong(r.getValue(family, qualifier));
        assertEquals(value, theStoredVal);
        // after flush
        LOG.info("Verification after flush...");
        TEST_UTIL.getAdmin().flush(tableName);
        r = ht.get(g);
        theStoredVal = Bytes.toLong(r.getValue(family, qualifier));
        assertEquals(value, theStoredVal);
    } finally {
        if (ht != null)
            ht.close();
        if (zkw != null)
            zkw.close();
    }
}