Examples with MeshProber org.voltdb.probe.MeshProber used on opensource projects

Example 6 with MeshProber

use of org.voltdb.probe.MeshProber in project voltdb by VoltDB.

the class TestHostMessenger method testProbedConfigMismatchCrash.

@Test
public void testProbedConfigMismatchCrash() throws Exception {
    VoltDB.ignoreCrash = true;
    MeshProber.Builder jcb = MeshProber.builder().coordinators(coordinators(2)).hostCount(3).startAction(StartAction.PROBE).nodeState(NodeState.INITIALIZING).kfactor(1).paused(false).bare(true);
    MeshProber jc1 = jcb.build();
    MeshProber jc2 = jcb.configHash(new UUID(-2L, -2L)).build();
    assertNotSame(jc1.getConfigHash(), jc2.getConfigHash());
    HostMessenger hm1 = createHostMessenger(0, jcb.prober(jc1).build(), false);
    HostMessenger hm2 = createHostMessenger(1, jcb.prober(jc1).build(), false);
    HostMessenger hm3 = createHostMessenger(2, jcb.prober(jc2).build(), false);
    final AtomicReference<Exception> exception = new AtomicReference<Exception>();
    HostMessengerThread hm1Start = new HostMessengerThread(hm1, exception);
    HostMessengerThread hm2Start = new HostMessengerThread(hm2, exception);
    hm1Start.start();
    hm2Start.start();
    hm1Start.join();
    hm2Start.join();
    if (exception.get() != null) {
        fail(exception.get().toString());
    }
    try {
        hm3.start();
        fail("did not crash on whole cluster rejoin attempt");
    } catch (AssertionError pass) {
        assertTrue(VoltDB.wasCrashCalled);
        assertTrue(VoltDB.crashMessage.contains("deployment options that do not match"));
    }
}

Example 7 with MeshProber

use of org.voltdb.probe.MeshProber in project voltdb by VoltDB.

the class TestHostMessenger method testMultipleMismatchesCrash.

@Test
public void testMultipleMismatchesCrash() throws Exception {
    VoltDB.ignoreCrash = true;
    MeshProber.Builder jcb = MeshProber.builder().coordinators(coordinators(2)).hostCount(3).startAction(StartAction.PROBE).nodeState(NodeState.INITIALIZING).kfactor(1).paused(false).bare(true);
    MeshProber jc1 = jcb.build();
    MeshProber jc2 = jcb.startAction(StartAction.CREATE).configHash(new UUID(-2L, -2L)).hostCount(4).build();
    assertNotSame(jc1.getStartAction(), jc2.getStartAction());
    assertNotSame(jc1.getHostCount(), jc2.getHostCount());
    assertNotSame(jc1.getConfigHash(), jc2.getConfigHash());
    HostMessenger hm1 = createHostMessenger(0, jcb.prober(jc1).build(), false);
    HostMessenger hm2 = createHostMessenger(1, jcb.prober(jc1).build(), false);
    HostMessenger hm3 = createHostMessenger(2, jcb.prober(jc2).build(), false);
    final AtomicReference<Exception> exception = new AtomicReference<Exception>();
    HostMessengerThread hm1Start = new HostMessengerThread(hm1, exception);
    HostMessengerThread hm2Start = new HostMessengerThread(hm2, exception);
    hm1Start.start();
    hm2Start.start();
    hm1Start.join();
    hm2Start.join();
    if (exception.get() != null) {
        fail(exception.get().toString());
    }
    try {
        hm3.start();
        fail("did not crash on whole cluster rejoin attempt");
    } catch (AssertionError pass) {
        assertTrue(VoltDB.wasCrashCalled);
        assertTrue(VoltDB.crashMessage.contains("Start action CREATE does not match PROBE"));
        assertTrue(VoltDB.crashMessage.contains("Mismatched host count"));
        assertTrue(VoltDB.crashMessage.contains("deployment options that do not match"));
    }
}

Example 8 with MeshProber

use of org.voltdb.probe.MeshProber in project voltdb by VoltDB.

the class TestHostMessenger method testProbedTerminusNonceMismatchCrash.

@Test
public void testProbedTerminusNonceMismatchCrash() throws Exception {
    VoltDB.ignoreCrash = true;
    MeshProber.Builder jcb = MeshProber.builder().coordinators(coordinators(2)).hostCount(3).startAction(StartAction.PROBE).nodeState(NodeState.INITIALIZING).kfactor(1).paused(false).bare(true).terminusNonce("uno");
    MeshProber jc1 = jcb.build();
    MeshProber jc2 = jcb.terminusNonce("due").build();
    assertNotSame(jc1.getTerminusNonce(), jc2.getTerminusNonce());
    HostMessenger hm1 = createHostMessenger(0, jcb.prober(jc1).build(), false);
    HostMessenger hm2 = createHostMessenger(1, jcb.prober(jc1).build(), false);
    HostMessenger hm3 = createHostMessenger(2, jcb.prober(jc2).build(), false);
    final AtomicReference<Exception> exception = new AtomicReference<Exception>();
    HostMessengerThread hm1Start = new HostMessengerThread(hm1, exception);
    HostMessengerThread hm2Start = new HostMessengerThread(hm2, exception);
    hm1Start.start();
    hm2Start.start();
    hm1Start.join();
    hm2Start.join();
    if (exception.get() != null) {
        fail(exception.get().toString());
    }
    try {
        hm3.start();
        fail("did not crash on whole cluster rejoin attempt");
    } catch (AssertionError pass) {
        assertTrue(VoltDB.wasCrashCalled);
        assertTrue(VoltDB.crashMessage.contains("have different startup snapshots nonces"));
    }
}

Example 9 with MeshProber

use of org.voltdb.probe.MeshProber in project voltdb by VoltDB.

the class TestHostMessenger method testSingleProbedCreateHost.

@Test
public void testSingleProbedCreateHost() throws Exception {
    MeshProber jc = MeshProber.builder().coordinators(coordinators(1)).startAction(StartAction.PROBE).nodeState(NodeState.INITIALIZING).bare(true).build();
    HostMessenger hm = createHostMessenger(0, jc, true);
    Determination dtm = prober(hm).waitForDetermination();
    assertEquals(StartAction.CREATE, dtm.startAction);
    assertEquals(1, dtm.hostCount);
    Mailbox m1 = hm.createMailbox();
    SiteMailbox sm = new SiteMailbox(hm, (-2L << 32));
    hm.createMailbox(sm.getHSId(), sm);
    sm.send(m1.getHSId(), new LocalObjectMessage(null));
    m1.send(sm.getHSId(), new LocalObjectMessage(null));
    LocalObjectMessage lom = (LocalObjectMessage) m1.recv();
    assertEquals(lom.m_sourceHSId, sm.getHSId());
    lom = (LocalObjectMessage) sm.recv();
    assertEquals(lom.m_sourceHSId, m1.getHSId());
}

Example 10 with MeshProber

use of org.voltdb.probe.MeshProber in project voltdb by VoltDB.

the class RealVoltDB method buildClusterMesh.

/**
     * Start the voltcore HostMessenger. This joins the node
     * to the existing cluster. In the non rejoin case, this
     * function will return when the mesh is complete. If
     * rejoining, it will return when the node and agreement
     * site are synched to the existing cluster.
     */
MeshProber.Determination buildClusterMesh(ReadDeploymentResults readDepl) {
    final boolean bareAtStartup = m_config.m_forceVoltdbCreate || pathsWithRecoverableArtifacts(readDepl.deployment).isEmpty();
    setBare(bareAtStartup);
    final Supplier<Integer> hostCountSupplier = new Supplier<Integer>() {

        @Override
        public Integer get() {
            return m_clusterSettings.get().hostcount();
        }
    };
    ClusterType clusterType = readDepl.deployment.getCluster();
    MeshProber criteria = MeshProber.builder().coordinators(m_config.m_coordinators).versionChecker(m_versionChecker).enterprise(m_config.m_isEnterprise).startAction(m_config.m_startAction).bare(bareAtStartup).configHash(CatalogUtil.makeDeploymentHashForConfig(readDepl.deploymentBytes)).hostCountSupplier(hostCountSupplier).kfactor(clusterType.getKfactor()).paused(m_config.m_isPaused).nodeStateSupplier(m_statusTracker.getNodeStateSupplier()).addAllowed(m_config.m_enableAdd).safeMode(m_config.m_safeMode).terminusNonce(getTerminusNonce()).missingHostCount(m_config.m_missingHostCount).build();
    HostAndPort hostAndPort = criteria.getLeader();
    String hostname = hostAndPort.getHostText();
    int port = hostAndPort.getPort();
    org.voltcore.messaging.HostMessenger.Config hmconfig;
    hmconfig = new org.voltcore.messaging.HostMessenger.Config(hostname, port);
    if (m_config.m_placementGroup != null) {
        hmconfig.group = m_config.m_placementGroup;
    }
    hmconfig.internalPort = m_config.m_internalPort;
    hmconfig.internalInterface = m_config.m_internalInterface;
    hmconfig.zkInterface = m_config.m_zkInterface;
    hmconfig.deadHostTimeout = m_config.m_deadHostTimeoutMS;
    hmconfig.factory = new VoltDbMessageFactory();
    hmconfig.coreBindIds = m_config.m_networkCoreBindings;
    hmconfig.acceptor = criteria;
    hmconfig.localSitesCount = m_config.m_sitesperhost;
    m_messenger = new org.voltcore.messaging.HostMessenger(hmconfig, this);
    hostLog.info(String.format("Beginning inter-node communication on port %d.", m_config.m_internalPort));
    try {
        m_messenger.start();
    } catch (Exception e) {
        VoltDB.crashLocalVoltDB(e.getMessage(), true, e);
    }
    VoltZK.createPersistentZKNodes(m_messenger.getZK());
    // Use the host messenger's hostId.
    m_myHostId = m_messenger.getHostId();
    hostLog.info(String.format("Host id of this node is: %d", m_myHostId));
    consoleLog.info(String.format("Host id of this node is: %d", m_myHostId));
    MeshProber.Determination determination = criteria.waitForDetermination();
    // paused is determined in the mesh formation exchanged
    if (determination.paused) {
        m_messenger.pause();
    } else {
        m_messenger.unpause();
    }
    // leader and we're rejoining, this is clearly bad.
    if (m_myHostId == 0 && determination.startAction.doesJoin()) {
        VoltDB.crashLocalVoltDB("Unable to rejoin a node to itself.  " + "Please check your command line and start action and try again.", false, null);
    }
    // load or store settings form/to zookeeper
    if (determination.startAction.doesJoin()) {
        m_clusterSettings.load(m_messenger.getZK());
        m_clusterSettings.get().store();
    } else if (m_myHostId == 0) {
        m_clusterSettings.store(m_messenger.getZK());
    }
    m_clusterCreateTime = m_messenger.getInstanceId().getTimestamp();
    return determination;
}