Examples with ClusterMetrics org.apache.hadoop.hbase.ClusterMetrics used on opensource projects

Example 31 with ClusterMetrics

use of org.apache.hadoop.hbase.ClusterMetrics in project hbase by apache.

the class TestRSGroupsBase method getNumServers.

// return the real number of region servers, excluding the master embedded region server in 2.0+
protected int getNumServers() throws IOException {
    ClusterMetrics status = ADMIN.getClusterMetrics(EnumSet.of(Option.MASTER, Option.LIVE_SERVERS));
    ServerName masterName = status.getMasterName();
    int count = 0;
    for (ServerName sn : status.getLiveServerMetrics().keySet()) {
        if (!sn.equals(masterName)) {
            count++;
        }
    }
    return count;
}

Example 32 with ClusterMetrics

use of org.apache.hadoop.hbase.ClusterMetrics in project hbase by apache.

the class TestRSGroupsBase method getTableServerRegionMap.

protected Map<TableName, Map<ServerName, List<String>>> getTableServerRegionMap() throws IOException {
    Map<TableName, Map<ServerName, List<String>>> map = Maps.newTreeMap();
    Admin admin = TEST_UTIL.getAdmin();
    ClusterMetrics metrics = admin.getClusterMetrics(EnumSet.of(ClusterMetrics.Option.SERVERS_NAME));
    for (ServerName serverName : metrics.getServersName()) {
        for (RegionInfo region : admin.getRegions(serverName)) {
            TableName tableName = region.getTable();
            map.computeIfAbsent(tableName, k -> new TreeMap<>()).computeIfAbsent(serverName, k -> new ArrayList<>()).add(region.getRegionNameAsString());
        }
    }
    return map;
}

Example 33 with ClusterMetrics

use of org.apache.hadoop.hbase.ClusterMetrics in project hbase by apache.

the class TestMasterFailover method testSimpleMasterFailover.

/**
 * Simple test of master failover.
 * <p>
 * Starts with three masters.  Kills a backup master.  Then kills the active
 * master.  Ensures the final master becomes active and we can still contact
 * the cluster.
 */
@Test
public void testSimpleMasterFailover() throws Exception {
    final int NUM_MASTERS = 3;
    final int NUM_RS = 3;
    // Start the cluster
    HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil();
    try {
        StartTestingClusterOption option = StartTestingClusterOption.builder().numMasters(NUM_MASTERS).numRegionServers(NUM_RS).numDataNodes(NUM_RS).build();
        TEST_UTIL.startMiniCluster(option);
        SingleProcessHBaseCluster cluster = TEST_UTIL.getHBaseCluster();
        // get all the master threads
        List<MasterThread> masterThreads = cluster.getMasterThreads();
        // wait for each to come online
        for (MasterThread mt : masterThreads) {
            assertTrue(mt.isAlive());
        }
        // verify only one is the active master and we have right number
        int numActive = 0;
        int activeIndex = -1;
        ServerName activeName = null;
        HMaster active = null;
        for (int i = 0; i < masterThreads.size(); i++) {
            if (masterThreads.get(i).getMaster().isActiveMaster()) {
                numActive++;
                activeIndex = i;
                active = masterThreads.get(activeIndex).getMaster();
                activeName = active.getServerName();
            }
        }
        assertEquals(1, numActive);
        assertEquals(NUM_MASTERS, masterThreads.size());
        LOG.info("Active master " + activeName);
        // Check that ClusterStatus reports the correct active and backup masters
        assertNotNull(active);
        ClusterMetrics status = active.getClusterMetrics();
        assertEquals(activeName, status.getMasterName());
        assertEquals(2, status.getBackupMasterNames().size());
        // attempt to stop one of the inactive masters
        int backupIndex = (activeIndex == 0 ? 1 : activeIndex - 1);
        HMaster master = cluster.getMaster(backupIndex);
        LOG.debug("\n\nStopping a backup master: " + master.getServerName() + "\n");
        cluster.stopMaster(backupIndex, false);
        cluster.waitOnMaster(backupIndex);
        // Verify still one active master and it's the same
        for (int i = 0; i < masterThreads.size(); i++) {
            if (masterThreads.get(i).getMaster().isActiveMaster()) {
                assertEquals(activeName, masterThreads.get(i).getMaster().getServerName());
                activeIndex = i;
                active = masterThreads.get(activeIndex).getMaster();
            }
        }
        assertEquals(1, numActive);
        assertEquals(2, masterThreads.size());
        int rsCount = masterThreads.get(activeIndex).getMaster().getClusterMetrics().getLiveServerMetrics().size();
        LOG.info("Active master " + active.getServerName() + " managing " + rsCount + " regions servers");
        assertEquals(3, rsCount);
        // wait for the active master to acknowledge loss of the backup from ZK
        final HMaster activeFinal = active;
        TEST_UTIL.waitFor(TimeUnit.MINUTES.toMillis(5), () -> activeFinal.getBackupMasters().size() == 1);
        // Check that ClusterStatus reports the correct active and backup masters
        assertNotNull(active);
        status = active.getClusterMetrics();
        assertEquals(activeName, status.getMasterName());
        assertEquals(1, status.getBackupMasterNames().size());
        // kill the active master
        LOG.debug("\n\nStopping the active master " + active.getServerName() + "\n");
        cluster.stopMaster(activeIndex, false);
        cluster.waitOnMaster(activeIndex);
        // wait for an active master to show up and be ready
        assertTrue(cluster.waitForActiveAndReadyMaster());
        LOG.debug("\n\nVerifying backup master is now active\n");
        // should only have one master now
        assertEquals(1, masterThreads.size());
        // and he should be active
        active = masterThreads.get(0).getMaster();
        assertNotNull(active);
        status = active.getClusterMetrics();
        ServerName masterName = status.getMasterName();
        assertNotNull(masterName);
        assertEquals(active.getServerName(), masterName);
        assertTrue(active.isActiveMaster());
        assertEquals(0, status.getBackupMasterNames().size());
        int rss = status.getLiveServerMetrics().size();
        LOG.info("Active master {} managing {} region servers", masterName.getServerName(), rss);
        assertEquals(3, rss);
    } finally {
        // Stop the cluster
        TEST_UTIL.shutdownMiniCluster();
    }
}

Example 34 with ClusterMetrics

use of org.apache.hadoop.hbase.ClusterMetrics in project hbase by apache.

the class TestMasterFailoverBalancerPersistence method testMasterFailoverBalancerPersistence.

/**
 * Test that if the master fails, the load balancer maintains its
 * state (running or not) when the next master takes over
 *
 * @throws Exception
 */
@Test
public void testMasterFailoverBalancerPersistence() throws Exception {
    // Start the cluster
    HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil();
    StartTestingClusterOption option = StartTestingClusterOption.builder().numMasters(3).build();
    TEST_UTIL.startMiniCluster(option);
    SingleProcessHBaseCluster cluster = TEST_UTIL.getHBaseCluster();
    assertTrue(cluster.waitForActiveAndReadyMaster());
    HMaster active = cluster.getMaster();
    // check that the balancer is on by default for the active master
    ClusterMetrics clusterStatus = active.getClusterMetrics();
    assertTrue(clusterStatus.getBalancerOn());
    active = killActiveAndWaitForNewActive(cluster);
    // ensure the load balancer is still running on new master
    clusterStatus = active.getClusterMetrics();
    assertTrue(clusterStatus.getBalancerOn());
    // turn off the load balancer
    active.balanceSwitch(false);
    // once more, kill active master and wait for new active master to show up
    active = killActiveAndWaitForNewActive(cluster);
    // ensure the load balancer is not running on the new master
    clusterStatus = active.getClusterMetrics();
    assertFalse(clusterStatus.getBalancerOn());
    // Stop the cluster
    TEST_UTIL.shutdownMiniCluster();
}

Example 35 with ClusterMetrics

use of org.apache.hadoop.hbase.ClusterMetrics in project hbase by apache.

the class TestReplicationStatusAfterLagging method testReplicationStatusAfterLagging.

@Test
public void testReplicationStatusAfterLagging() throws Exception {
    UTIL2.shutdownMiniHBaseCluster();
    restartSourceCluster(1);
    // add some values to cluster 1
    for (int i = 0; i < NB_ROWS_IN_BATCH; i++) {
        Put p = new Put(Bytes.toBytes("row" + i));
        p.addColumn(famName, Bytes.toBytes("col1"), Bytes.toBytes("val" + i));
        htable1.put(p);
    }
    UTIL2.startMiniHBaseCluster();
    Thread.sleep(10000);
    Admin hbaseAdmin = UTIL1.getAdmin();
    ServerName serverName = UTIL1.getHBaseCluster().getRegionServer(0).getServerName();
    ClusterMetrics metrics = hbaseAdmin.getClusterMetrics(EnumSet.of(Option.LIVE_SERVERS));
    List<ReplicationLoadSource> loadSources = metrics.getLiveServerMetrics().get(serverName).getReplicationLoadSourceList();
    assertEquals(1, loadSources.size());
    ReplicationLoadSource loadSource = loadSources.get(0);
    assertTrue(loadSource.hasEditsSinceRestart());
    assertTrue(loadSource.getTimestampOfLastShippedOp() > 0);
    assertEquals(0, loadSource.getReplicationLag());
}