Examples with HRegionServer org.apache.hadoop.hbase.regionserver.HRegionServer used on opensource projects

Example 81 with HRegionServer

use of org.apache.hadoop.hbase.regionserver.HRegionServer in project hbase by apache.

the class TestMasterObserver method testRegionTransitionOperations.

@Test
public void testRegionTransitionOperations() throws Exception {
    final TableName tableName = TableName.valueOf(name.getMethodName());
    SingleProcessHBaseCluster cluster = UTIL.getHBaseCluster();
    HMaster master = cluster.getMaster();
    MasterCoprocessorHost host = master.getMasterCoprocessorHost();
    CPMasterObserver cp = host.findCoprocessor(CPMasterObserver.class);
    cp.resetStates();
    Table table = UTIL.createMultiRegionTable(tableName, TEST_FAMILY);
    try (RegionLocator r = UTIL.getConnection().getRegionLocator(tableName)) {
        UTIL.waitUntilAllRegionsAssigned(tableName);
        List<HRegionLocation> regions = r.getAllRegionLocations();
        HRegionLocation firstGoodPair = null;
        for (HRegionLocation e : regions) {
            if (e.getServerName() != null) {
                firstGoodPair = e;
                break;
            }
        }
        assertNotNull("Found a non-null entry", firstGoodPair);
        LOG.info("Found " + firstGoodPair.toString());
        // Try to force a move
        Collection<ServerName> servers = master.getClusterMetrics().getLiveServerMetrics().keySet();
        String destName = null;
        String serverNameForFirstRegion = firstGoodPair.getServerName().toString();
        LOG.info("serverNameForFirstRegion=" + serverNameForFirstRegion);
        ServerName masterServerName = master.getServerName();
        boolean found = false;
        // Find server that is NOT carrying the first region
        for (ServerName info : servers) {
            LOG.info("ServerName=" + info);
            if (!serverNameForFirstRegion.equals(info.getServerName()) && !masterServerName.equals(info)) {
                destName = info.toString();
                found = true;
                break;
            }
        }
        assertTrue("Found server", found);
        LOG.info("Found " + destName);
        master.getMasterRpcServices().moveRegion(null, RequestConverter.buildMoveRegionRequest(firstGoodPair.getRegion().getEncodedNameAsBytes(), ServerName.valueOf(destName)));
        assertTrue("Coprocessor should have been called on region move", cp.wasMoveCalled());
        // make sure balancer is on
        master.balanceSwitch(true);
        assertTrue("Coprocessor should have been called on balance switch", cp.wasBalanceSwitchCalled());
        // turn balancer off
        master.balanceSwitch(false);
        // wait for assignments to finish, if any
        UTIL.waitUntilNoRegionsInTransition();
        // move half the open regions from RS 0 to RS 1
        HRegionServer rs = cluster.getRegionServer(0);
        byte[] destRS = Bytes.toBytes(cluster.getRegionServer(1).getServerName().toString());
        // Make sure no regions are in transition now
        UTIL.waitUntilNoRegionsInTransition();
        List<RegionInfo> openRegions = ProtobufUtil.getOnlineRegions(rs.getRSRpcServices());
        int moveCnt = openRegions.size() / 2;
        for (int i = 0; i < moveCnt; i++) {
            RegionInfo info = openRegions.get(i);
            if (!info.isMetaRegion()) {
                master.getMasterRpcServices().moveRegion(null, RequestConverter.buildMoveRegionRequest(openRegions.get(i).getEncodedNameAsBytes(), ServerName.valueOf(Bytes.toString(destRS))));
            }
        }
        // Make sure no regions are in transition now
        UTIL.waitUntilNoRegionsInTransition();
        // now trigger a balance
        master.balanceSwitch(true);
        master.balance();
        assertTrue("Coprocessor should be called on region rebalancing", cp.wasBalanceCalled());
    } finally {
        Admin admin = UTIL.getAdmin();
        admin.disableTable(tableName);
        deleteTable(admin, tableName);
    }
}

Example 82 with HRegionServer

use of org.apache.hadoop.hbase.regionserver.HRegionServer in project hbase by apache.

the class TestRegionServerCoprocessorExceptionWithAbort method testExceptionDuringInitialization.

@Test
public void testExceptionDuringInitialization() throws Exception {
    Configuration conf = TEST_UTIL.getConfiguration();
    // Let's fail fast.
    conf.setInt(HConstants.HBASE_CLIENT_RETRIES_NUMBER, 2);
    conf.setBoolean(CoprocessorHost.ABORT_ON_ERROR_KEY, true);
    conf.set(CoprocessorHost.REGION_COPROCESSOR_CONF_KEY, "");
    TEST_UTIL.startMiniCluster(2);
    try {
        SingleProcessHBaseCluster cluster = TEST_UTIL.getHBaseCluster();
        // Trigger one regionserver to fail as if it came up with a coprocessor
        // that fails during initialization
        final HRegionServer regionServer = cluster.getRegionServer(0);
        conf.set(CoprocessorHost.REGION_COPROCESSOR_CONF_KEY, FailedInitializationObserver.class.getName());
        regionServer.getRegionServerCoprocessorHost().loadSystemCoprocessors(conf, CoprocessorHost.REGION_COPROCESSOR_CONF_KEY);
        TEST_UTIL.waitFor(10000, 1000, new Predicate<Exception>() {

            @Override
            public boolean evaluate() throws Exception {
                return regionServer.isAborted();
            }
        });
    } finally {
        TEST_UTIL.shutdownMiniCluster();
    }
}

Example 83 with HRegionServer

use of org.apache.hadoop.hbase.regionserver.HRegionServer in project hbase by apache.

the class TestRegionServerCoprocessorExceptionWithAbort method testExceptionFromCoprocessorDuringPut.

@Test
public void testExceptionFromCoprocessorDuringPut() throws Exception {
    // set configure to indicate which cp should be loaded
    Configuration conf = TEST_UTIL.getConfiguration();
    // Let's fail fast.
    conf.setInt(HConstants.HBASE_CLIENT_RETRIES_NUMBER, 2);
    conf.set(CoprocessorHost.REGION_COPROCESSOR_CONF_KEY, BuggyRegionObserver.class.getName());
    conf.setBoolean(CoprocessorHost.ABORT_ON_ERROR_KEY, true);
    TEST_UTIL.startMiniCluster(2);
    try {
        // When we try to write to TEST_TABLE, the buggy coprocessor will
        // cause a NullPointerException, which will cause the regionserver (which
        // hosts the region we attempted to write to) to abort.
        final byte[] TEST_FAMILY = Bytes.toBytes("aaa");
        Table table = TEST_UTIL.createMultiRegionTable(TABLE_NAME, TEST_FAMILY);
        TEST_UTIL.waitUntilAllRegionsAssigned(TABLE_NAME);
        // Note which regionServer will abort (after put is attempted).
        final HRegionServer regionServer = TEST_UTIL.getRSForFirstRegionInTable(TABLE_NAME);
        try {
            final byte[] ROW = Bytes.toBytes("aaa");
            Put put = new Put(ROW);
            put.addColumn(TEST_FAMILY, ROW, ROW);
            table.put(put);
        } catch (IOException e) {
        // The region server is going to be aborted.
        // We may get an exception if we retry,
        // which is not guaranteed.
        }
        // Wait 10 seconds for the regionserver to abort: expected result is that
        // it will abort.
        boolean aborted = false;
        for (int i = 0; i < 10; i++) {
            aborted = regionServer.isAborted();
            if (aborted) {
                break;
            }
            try {
                Thread.sleep(1000);
            } catch (InterruptedException e) {
                fail("InterruptedException while waiting for regionserver " + "zk node to be deleted.");
            }
        }
        Assert.assertTrue("The region server should have aborted", aborted);
        table.close();
    } finally {
        TEST_UTIL.shutdownMiniCluster();
    }
}

Example 84 with HRegionServer

use of org.apache.hadoop.hbase.regionserver.HRegionServer in project hbase by apache.

the class TestRegionServerCoprocessorExceptionWithRemove method testExceptionFromCoprocessorDuringPut.

@Test
public void testExceptionFromCoprocessorDuringPut() throws IOException, InterruptedException {
    // Set watches on the zookeeper nodes for all of the regionservers in the
    // cluster. When we try to write to TEST_TABLE, the buggy coprocessor will
    // cause a NullPointerException, which will cause the regionserver (which
    // hosts the region we attempted to write to) to abort. In turn, this will
    // cause the nodeDeleted() method of the DeadRegionServer tracker to
    // execute, which will set the rsZKNodeDeleted flag to true, which will
    // pass this test.
    TableName TEST_TABLE = TableName.valueOf("observed_table");
    byte[] TEST_FAMILY = Bytes.toBytes("aaa");
    Table table = TEST_UTIL.createMultiRegionTable(TEST_TABLE, TEST_FAMILY);
    TEST_UTIL.waitUntilAllRegionsAssigned(TEST_TABLE);
    // Note which regionServer that should survive the buggy coprocessor's
    // prePut().
    HRegionServer regionServer = TEST_UTIL.getRSForFirstRegionInTable(TEST_TABLE);
    boolean threwIOE = false;
    try {
        final byte[] ROW = Bytes.toBytes("aaa");
        Put put = new Put(ROW);
        put.addColumn(TEST_FAMILY, ROW, ROW);
        table.put(put);
        // We may need two puts to reliably get an exception
        table.put(put);
    } catch (IOException e) {
        threwIOE = true;
    } finally {
        assertTrue("The regionserver should have thrown an exception", threwIOE);
    }
    // it will survive and not abort.
    for (int i = 0; i < 10; i++) {
        assertFalse(regionServer.isAborted());
        try {
            Thread.sleep(1000);
        } catch (InterruptedException e) {
            fail("InterruptedException while waiting for regionserver " + "zk node to be deleted.");
        }
    }
    table.close();
}

Example 85 with HRegionServer

use of org.apache.hadoop.hbase.regionserver.HRegionServer in project hbase by apache.

the class TestProcedurePriority method test.

@Test
public void test() throws Exception {
    RegionServerThread rsWithMetaThread = UTIL.getMiniHBaseCluster().getRegionServerThreads().stream().filter(t -> !t.getRegionServer().getRegions(TableName.META_TABLE_NAME).isEmpty()).findAny().get();
    HRegionServer rsNoMeta = UTIL.getOtherRegionServer(rsWithMetaThread.getRegionServer());
    FAIL = true;
    UTIL.getMiniHBaseCluster().killRegionServer(rsNoMeta.getServerName());
    // wait until all the worker thread are stuck, which means that the stuck checker will start to
    // add new worker thread.
    ProcedureExecutor<?> executor = UTIL.getMiniHBaseCluster().getMaster().getMasterProcedureExecutor();
    UTIL.waitFor(60000, new ExplainingPredicate<Exception>() {

        @Override
        public boolean evaluate() throws Exception {
            return executor.getWorkerThreadCount() > CORE_POOL_SIZE;
        }

        @Override
        public String explainFailure() throws Exception {
            return "Stuck checker does not add new worker thread";
        }
    });
    UTIL.getMiniHBaseCluster().killRegionServer(rsWithMetaThread.getRegionServer().getServerName());
    rsWithMetaThread.join();
    FAIL = false;
    // verify that the cluster is back
    UTIL.waitUntilNoRegionsInTransition(480000);
    for (int i = 0; i < TABLE_COUNT; i++) {
        try (Table table = UTIL.getConnection().getTable(TableName.valueOf(TABLE_NAME_PREFIX + i))) {
            table.put(new Put(Bytes.toBytes(i)).addColumn(CF, CQ, Bytes.toBytes(i)));
        }
    }
    UTIL.waitFor(60000, new ExplainingPredicate<Exception>() {

        @Override
        public boolean evaluate() throws Exception {
            return executor.getWorkerThreadCount() == CORE_POOL_SIZE;
        }

        @Override
        public String explainFailure() throws Exception {
            return "The new workers do not timeout";
        }
    });
}