Examples with ParameterServer com.tencent.angel.ps.ParameterServer used on opensource projects

Example 1 with ParameterServer

use of com.tencent.angel.ps.ParameterServer in project angel by Tencent.

the class PSManagerTest method testPSError.

@Test
public void testPSError() throws Exception {
    try {
        int heartbeatInterval = LocalClusterContext.get().getConf().getInt(AngelConf.ANGEL_PS_HEARTBEAT_INTERVAL_MS, AngelConf.DEFAULT_ANGEL_PS_HEARTBEAT_INTERVAL_MS);
        AngelApplicationMaster angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        ParameterServerManager psManager = angelAppMaster.getAppContext().getParameterServerManager();
        AMParameterServer amPs = psManager.getParameterServer(psId);
        PSAttempt psAttempt0 = amPs.getPSAttempt(psAttempt0Id);
        ParameterServer ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
        Worker worker = LocalClusterContext.get().getWorker(worker0Attempt0Id).getWorker();
        int w1Id = angelAppMaster.getAppContext().getMatrixMetaManager().getMatrix("w1").getId();
        int w2Id = angelAppMaster.getAppContext().getMatrixMetaManager().getMatrix("w2").getId();
        Location masterLoc = LocalClusterContext.get().getMaster().getAppMaster().getAppContext().getMasterService().getLocation();
        TConnection connection = TConnectionManager.getConnection(ps.getConf());
        MasterProtocol master = connection.getMasterService(masterLoc.getIp(), masterLoc.getPort());
        int task0Iteration = 2;
        int task1Iteration = 1;
        int task0w1Clock = 10;
        int task0w2Clock = 20;
        int task1w1Clock = 9;
        int task1w2Clock = 19;
        int w1Clock = (task0w1Clock < task1w1Clock) ? task0w1Clock : task1w1Clock;
        int w2Clock = (task0w2Clock < task1w2Clock) ? task0w2Clock : task1w2Clock;
        TaskContext task0Context = worker.getTaskManager().getRunningTask().get(task0Id).getTaskContext().getContext();
        TaskContext task1Context = worker.getTaskManager().getRunningTask().get(task1Id).getTaskContext().getContext();
        master.taskIteration(null, TaskIterationRequest.newBuilder().setIteration(task0Iteration).setTaskId(ProtobufUtil.convertToIdProto(task0Id)).build());
        master.taskIteration(null, TaskIterationRequest.newBuilder().setIteration(task1Iteration).setTaskId(ProtobufUtil.convertToIdProto(task1Id)).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task0Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w1Id).setClock(task0w1Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task0Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w2Id).setClock(task0w2Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task1Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w1Id).setClock(task1w1Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task1Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w2Id).setClock(task1w2Clock).build()).build());
        assertEquals(amPs.getMaxAttempts(), 4);
        PSAttemptId psAttempt1Id = new PSAttemptId(psId, 1);
        PSAttemptId psAttempt2Id = new PSAttemptId(psId, 2);
        PSAttemptId psAttempt3Id = new PSAttemptId(psId, 3);
        // attempt 0
        ps.stop(-1);
        PSErrorRequest request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt0Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        PSAttempt psAttempt1 = amPs.getPSAttempt(psAttempt1Id);
        assertTrue(psAttempt1 != null);
        assertEquals(psAttempt0.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(psAttempt1.getInternalState(), PSAttemptStateInternal.RUNNING);
        assertEquals(amPs.getState(), AMParameterServerState.RUNNING);
        assertEquals(amPs.getNextAttemptNumber(), 2);
        assertEquals(amPs.getRunningAttemptId(), psAttempt1Id);
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 2);
        List<String> diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 1);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        ps = LocalClusterContext.get().getPS(psAttempt1Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock, w2Clock);
        MatrixClient w1Task0Client = worker.getPSAgent().getMatrixClient("w1", 0);
        MatrixClient w1Task1Client = worker.getPSAgent().getMatrixClient("w1", 1);
        int matrixW1Id = w1Task0Client.getMatrixId();
        int[] delta = new int[100000];
        for (int i = 0; i < 100000; i++) {
            delta[i] = 2;
        }
        IntIntVector deltaVec = new IntIntVector(100000, new IntIntDenseVectorStorage(delta));
        deltaVec.setMatrixId(matrixW1Id);
        deltaVec.setRowId(0);
        w1Task0Client.increment(deltaVec);
        deltaVec = new IntIntVector(100000, new IntIntDenseVectorStorage(delta));
        deltaVec.setMatrixId(matrixW1Id);
        deltaVec.setRowId(0);
        w1Task1Client.increment(deltaVec);
        w1Task0Client.clock().get();
        w1Task1Client.clock().get();
        ps = LocalClusterContext.get().getPS(psAttempt1Id).getPS();
        int snapshotInterval = LocalClusterContext.get().getConf().getInt(AngelConf.ANGEL_PS_BACKUP_INTERVAL_MS, AngelConf.DEFAULT_ANGEL_PS_BACKUP_INTERVAL_MS);
        Thread.sleep(snapshotInterval * 2);
        // attempt1
        ps.stop(-1);
        request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt1Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        PSAttempt psAttempt2 = amPs.getPSAttempt(psAttempt2Id);
        assertTrue(psAttempt2 != null);
        assertEquals(psAttempt1.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(psAttempt2.getInternalState(), PSAttemptStateInternal.RUNNING);
        assertEquals(amPs.getState(), AMParameterServerState.RUNNING);
        assertEquals(amPs.getNextAttemptNumber(), 3);
        assertEquals(amPs.getRunningAttemptId(), psAttempt2Id);
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 3);
        diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 2);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(1), psAttempt1Id + " failed due to: out of memory");
        ps = LocalClusterContext.get().getPS(psAttempt2Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock + 1, w2Clock);
        assertEquals(sum((IntIntVector) w1Task0Client.getRow(0)), 400000);
        // attempt1
        ps.stop(-1);
        request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt2Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        PSAttempt psAttempt3 = amPs.getPSAttempt(psAttempt3Id);
        assertTrue(psAttempt3 != null);
        assertEquals(psAttempt2.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(psAttempt3.getInternalState(), PSAttemptStateInternal.RUNNING);
        assertEquals(amPs.getState(), AMParameterServerState.RUNNING);
        assertEquals(amPs.getNextAttemptNumber(), 4);
        assertEquals(amPs.getRunningAttemptId(), psAttempt3Id);
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 4);
        diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 3);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(1), psAttempt1Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(2), psAttempt2Id + " failed due to: out of memory");
        ps = LocalClusterContext.get().getPS(psAttempt3Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock + 1, w2Clock);
        ps.stop(-1);
        request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt3Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        assertEquals(psAttempt3.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(amPs.getState(), AMParameterServerState.FAILED);
        assertEquals(angelAppMaster.getAppContext().getApp().getExternAppState(), AppState.FAILED);
        assertEquals(amPs.getNextAttemptNumber(), 4);
        assertNull(amPs.getRunningAttemptId());
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 4);
        diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 4);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(1), psAttempt1Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(2), psAttempt2Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(3), psAttempt3Id + " failed due to: out of memory");
    } catch (Exception x) {
        LOG.error("run testPSError failed ", x);
        throw x;
    }
}

Example 2 with ParameterServer

use of com.tencent.angel.ps.ParameterServer in project angel by Tencent.

the class PSManagerTest method testPSReport.

@Test
public void testPSReport() throws Exception {
    try {
        ParameterServer ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
        Location masterLoc = ps.getMasterLocation();
        TConnection connection = TConnectionManager.getConnection(ps.getConf());
        MasterProtocol master = connection.getMasterService(masterLoc.getIp(), masterLoc.getPort());
        PSReportRequest.Builder builder = PSReportRequest.newBuilder();
        builder.setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt0Id));
        Pair.Builder pairBuilder = Pair.newBuilder();
        pairBuilder.setKey("ps_key1");
        pairBuilder.setValue("100");
        builder.addMetrics(pairBuilder.build());
        pairBuilder.setKey("ps_key2");
        pairBuilder.setValue("200");
        builder.addMetrics(pairBuilder.build());
        MatrixReportProto.Builder matrixBuilder = MatrixReportProto.newBuilder();
        ConcurrentHashMap<Integer, ServerMatrix> matrixIdMap = ps.getMatrixStorageManager().getMatrices();
        for (Entry<Integer, ServerMatrix> matrixEntry : matrixIdMap.entrySet()) {
            builder.addMatrixReports((matrixBuilder.setMatrixId(matrixEntry.getKey()).setMatrixName(matrixEntry.getValue().getName())));
        }
        PSReportResponse response = master.psReport(null, builder.build());
        assertEquals(response.getPsCommand(), PSCommandProto.PSCOMMAND_OK);
        assertEquals(response.getNeedCreateMatricesCount(), 0);
        assertEquals(response.getNeedReleaseMatrixIdsCount(), 0);
        AngelApplicationMaster angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        ParameterServerManager psManager = angelAppMaster.getAppContext().getParameterServerManager();
        AMParameterServer amPs = psManager.getParameterServer(psId);
        PSAttempt psAttempt = amPs.getPSAttempt(psAttempt0Id);
        Map<String, String> metrices = psAttempt.getMetrices();
        assertTrue(metrices.get("ps_key1").equals("100"));
        assertTrue(metrices.get("ps_key2").equals("200"));
        PSAttemptId psAttempt1Id = new PSAttemptId(psId, 1);
        builder.setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt1Id));
        response = master.psReport(null, builder.build());
        assertEquals(response.getPsCommand(), PSCommandProto.PSCOMMAND_SHUTDOWN);
    } catch (Exception x) {
        LOG.error("run testPSReport failed ", x);
        throw x;
    }
}

Example 3 with ParameterServer

use of com.tencent.angel.ps.ParameterServer in project angel by Tencent.

the class MatrixMetaManagerTest method testCreateMatrix.

@Test
public void testCreateMatrix() throws Exception {
    try {
        LOG.info("===========================testCreateMatrix===============================");
        Worker worker = LocalClusterContext.get().getWorker(worker0Attempt0Id).getWorker();
        MasterClient masterClient = worker.getPSAgent().getMasterClient();
        int w3Id = -1;
        int w4Id = -1;
        // add matrix
        MatrixContext mMatrix = new MatrixContext();
        mMatrix.setName("w3");
        mMatrix.setRowNum(1);
        mMatrix.setColNum(100000);
        mMatrix.setMaxRowNumInBlock(1);
        mMatrix.setMaxColNumInBlock(50000);
        mMatrix.setRowType(RowType.T_DOUBLE_DENSE);
        mMatrix.set(MatrixConf.MATRIX_OPLOG_ENABLEFILTER, "false");
        mMatrix.set(MatrixConf.MATRIX_HOGWILD, "true");
        mMatrix.set(MatrixConf.MATRIX_AVERAGE, "false");
        mMatrix.set(MatrixConf.MATRIX_OPLOG_TYPE, RowType.T_DOUBLE_DENSE.name());
        masterClient.createMatrix(mMatrix, 10000);
        mMatrix.setName("w4");
        mMatrix.setRowNum(1);
        mMatrix.setColNum(100000);
        mMatrix.setMaxRowNumInBlock(1);
        mMatrix.setMaxColNumInBlock(50000);
        mMatrix.setRowType(RowType.T_DOUBLE_DENSE);
        mMatrix.set(MatrixConf.MATRIX_OPLOG_ENABLEFILTER, "false");
        mMatrix.set(MatrixConf.MATRIX_HOGWILD, "true");
        mMatrix.set(MatrixConf.MATRIX_AVERAGE, "false");
        mMatrix.set(MatrixConf.MATRIX_OPLOG_TYPE, RowType.T_DOUBLE_DENSE.name());
        masterClient.createMatrix(mMatrix, 10000);
        MatrixMeta w3Meta = worker.getPSAgent().getMatrixMetaManager().getMatrixMeta("w3");
        MatrixMeta w4Meta = worker.getPSAgent().getMatrixMetaManager().getMatrixMeta("w4");
        assertEquals(w3Meta.getRowNum(), 1);
        assertEquals(w3Meta.getColNum(), 100000);
        assertEquals(w3Meta.getRowType(), RowType.T_DOUBLE_DENSE);
        assertEquals(w4Meta.getRowNum(), 1);
        assertEquals(w4Meta.getColNum(), 100000);
        assertEquals(w4Meta.getRowType(), RowType.T_DOUBLE_DENSE);
        w3Id = w3Meta.getId();
        w4Id = w4Meta.getId();
        AngelApplicationMaster angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        assertTrue(angelAppMaster != null);
        AMMatrixMetaManager matrixMetaManager = angelAppMaster.getAppContext().getMatrixMetaManager();
        MatrixMeta matrixw3Proto = matrixMetaManager.getMatrix("w3");
        MatrixMeta matrixw4Proto = matrixMetaManager.getMatrix("w4");
        assertNotNull(matrixw3Proto);
        assertNotNull(matrixw4Proto);
        assertEquals(matrixw3Proto.getRowNum(), 1);
        assertEquals(matrixw3Proto.getColNum(), 100000);
        assertEquals(matrixw3Proto.getPartitionMetas().size(), 2);
        Map<Integer, PartitionMeta> w3Parts = matrixw3Proto.getPartitionMetas();
        assertEquals(w3Parts.get(0).getPss().get(0), psId);
        assertEquals(w3Parts.get(0).getPartId(), 0);
        assertEquals(w3Parts.get(0).getStartRow(), 0);
        assertEquals(w3Parts.get(0).getEndRow(), 1);
        assertEquals(w3Parts.get(0).getStartCol(), 0);
        assertEquals(w3Parts.get(0).getEndCol(), 50000);
        assertEquals(w3Parts.get(1).getPartId(), 1);
        assertEquals(w3Parts.get(1).getStartRow(), 0);
        assertEquals(w3Parts.get(1).getEndRow(), 1);
        assertEquals(w3Parts.get(1).getStartCol(), 50000);
        assertEquals(w3Parts.get(1).getEndCol(), 100000);
        Map<Integer, PartitionMeta> w4Parts = matrixw4Proto.getPartitionMetas();
        assertEquals(w4Parts.get(0).getPss().get(0), psId);
        assertEquals(w4Parts.get(0).getPartId(), 0);
        assertEquals(w4Parts.get(0).getStartRow(), 0);
        assertEquals(w4Parts.get(0).getEndRow(), 1);
        assertEquals(w4Parts.get(0).getStartCol(), 0);
        assertEquals(w4Parts.get(0).getEndCol(), 50000);
        assertEquals(w4Parts.get(1).getPartId(), 1);
        assertEquals(w4Parts.get(1).getStartRow(), 0);
        assertEquals(w4Parts.get(1).getEndRow(), 1);
        assertEquals(w4Parts.get(1).getStartCol(), 50000);
        assertEquals(w4Parts.get(1).getEndCol(), 100000);
        ParameterServer ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
        PSMatrixMetaManager matrixPartManager = ps.getMatrixMetaManager();
        PartitionMeta w3Part0 = matrixPartManager.getPartMeta(w3Id, 0);
        PartitionMeta w3Part1 = matrixPartManager.getPartMeta(w3Id, 1);
        assertTrue(w3Part0 != null);
        assertTrue(w3Part1 != null);
        assertEquals(w3Part0.getPartitionKey().getStartRow(), 0);
        assertEquals(w3Part0.getPartitionKey().getEndRow(), 1);
        assertEquals(w3Part0.getPartitionKey().getStartCol(), 0);
        assertEquals(w3Part0.getPartitionKey().getEndCol(), 50000);
        assertEquals(w3Part1.getPartitionKey().getStartRow(), 0);
        assertEquals(w3Part1.getPartitionKey().getEndRow(), 1);
        assertEquals(w3Part1.getPartitionKey().getStartCol(), 50000);
        assertEquals(w3Part1.getPartitionKey().getEndCol(), 100000);
        PartitionMeta w4Part0 = matrixPartManager.getPartMeta(w4Id, 0);
        PartitionMeta w4Part1 = matrixPartManager.getPartMeta(w4Id, 1);
        assertTrue(w4Part0 != null);
        assertTrue(w4Part1 != null);
        assertEquals(w4Part0.getPartitionKey().getStartRow(), 0);
        assertEquals(w4Part0.getPartitionKey().getEndRow(), 1);
        assertEquals(w4Part0.getPartitionKey().getStartCol(), 0);
        assertEquals(w4Part0.getPartitionKey().getEndCol(), 50000);
        assertEquals(w4Part1.getPartitionKey().getStartRow(), 0);
        assertEquals(w4Part1.getPartitionKey().getEndRow(), 1);
        assertEquals(w4Part1.getPartitionKey().getStartCol(), 50000);
        assertEquals(w4Part1.getPartitionKey().getEndCol(), 100000);
        MatrixClient w4ClientForTask0 = worker.getPSAgent().getMatrixClient("w4", 0);
        MatrixClient w4ClientForTask1 = worker.getPSAgent().getMatrixClient("w4", 1);
        TaskContext task0Context = w4ClientForTask0.getTaskContext();
        TaskContext task1Context = w4ClientForTask1.getTaskContext();
        double[] delta = new double[100000];
        for (int i = 0; i < delta.length; i++) {
            delta[i] = 1.0;
        }
        int iterIndex = 0;
        while (iterIndex < 5) {
            IntDoubleVector row1 = (IntDoubleVector) w4ClientForTask0.getRow(0);
            double sum1 = sum(row1.getStorage().getValues());
            LOG.info("taskid=" + task0Context.getIndex() + ", matrixId=" + w4ClientForTask0.getMatrixId() + ", rowIndex=0, local row sum=" + sum1);
            IntDoubleVector deltaRow1 = new IntDoubleVector(delta.length, new IntDoubleDenseVectorStorage(delta));
            deltaRow1.setMatrixId(w4ClientForTask0.getMatrixId());
            deltaRow1.setRowId(0);
            w4ClientForTask0.increment(deltaRow1);
            w4ClientForTask0.clock().get();
            task0Context.increaseEpoch();
            IntDoubleVector row2 = (IntDoubleVector) w4ClientForTask1.getRow(0);
            double sum2 = sum(row2.getStorage().getValues());
            LOG.info("taskid=" + task1Context.getIndex() + ", matrixId=" + w4ClientForTask1.getMatrixId() + ", rowIndex=1, local row sum=" + sum2);
            IntDoubleVector deltaRow2 = new IntDoubleVector(delta.length, new IntDoubleDenseVectorStorage(delta));
            deltaRow2.setMatrixId(w4ClientForTask1.getMatrixId());
            deltaRow2.setRowId(0);
            w4ClientForTask1.increment(deltaRow2);
            w4ClientForTask1.clock().get();
            task1Context.increaseEpoch();
            iterIndex++;
        }
        AMTaskManager amTaskManager = angelAppMaster.getAppContext().getTaskManager();
        AMTask amTask0 = amTaskManager.getTask(task0Id);
        AMTask amTask1 = amTaskManager.getTask(task1Id);
        assertEquals(amTask0.getIteration(), 5);
        assertEquals(amTask1.getIteration(), 5);
        Int2IntOpenHashMap task0MatrixClocks = amTask0.getMatrixClocks();
        assertEquals(task0MatrixClocks.size(), 1);
        assertEquals(task0MatrixClocks.get(w4Id), 5);
        Int2IntOpenHashMap task1MatrixClocks = amTask1.getMatrixClocks();
        assertEquals(task1MatrixClocks.size(), 1);
        assertEquals(task1MatrixClocks.get(w4Id), 5);
        IntDoubleVector row1 = (IntDoubleVector) w4ClientForTask0.getRow(0);
        double sum1 = sum(row1.getStorage().getValues());
        assertEquals(sum1, 1000000.0, 0.000001);
        IntDoubleVector row2 = (IntDoubleVector) w4ClientForTask1.getRow(0);
        double sum2 = sum(row2.getStorage().getValues());
        assertEquals(sum2, 1000000.0, 0.000001);
        masterClient.releaseMatrix(w3Meta.getName());
        Thread.sleep(10000);
        matrixw3Proto = matrixMetaManager.getMatrix("w3");
        assertTrue(matrixw3Proto == null);
        MatrixStorageManager matrixStorageManager = LocalClusterContext.get().getPS(psAttempt0Id).getPS().getMatrixStorageManager();
        ServerMatrix sw3 = matrixStorageManager.getMatrix(w3Id);
        assertTrue(sw3 == null);
        w4ClientForTask0.clock().get();
        w4ClientForTask1.clock().get();
        row1 = (IntDoubleVector) w4ClientForTask0.getRow(0);
        sum1 = sum(row1.getStorage().getValues());
        assertEquals(sum1, 1000000.0, 0.000001);
        row2 = (IntDoubleVector) w4ClientForTask1.getRow(0);
        sum2 = sum(row2.getStorage().getValues());
        assertEquals(sum2, 1000000.0, 0.000001);
    } catch (Exception x) {
        LOG.error("run testCreateMatrix failed ", x);
        throw x;
    }
}

Example 4 with ParameterServer

use of com.tencent.angel.ps.ParameterServer in project angel by Tencent.

the class MasterRecoverTest method testMasterRecover.

@SuppressWarnings("unchecked")
@Test
public void testMasterRecover() throws Exception {
    try {
        ApplicationAttemptId appAttempt1Id = ApplicationAttemptId.newInstance(LocalClusterContext.get().getAppId(), 1);
        ApplicationAttemptId appAttempt2Id = ApplicationAttemptId.newInstance(LocalClusterContext.get().getAppId(), 2);
        AngelApplicationMaster angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        LOG.info("angelAppMaster.getAppContext().getApplicationAttemptId()=" + angelAppMaster.getAppContext().getApplicationAttemptId());
        assertEquals(angelAppMaster.getAppContext().getApplicationAttemptId(), appAttempt1Id);
        ParameterServerManager psManager = angelAppMaster.getAppContext().getParameterServerManager();
        ParameterServer ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
        int w1Id = angelAppMaster.getAppContext().getMatrixMetaManager().getMatrix("w1").getId();
        int w2Id = angelAppMaster.getAppContext().getMatrixMetaManager().getMatrix("w2").getId();
        Location masterLoc = LocalClusterContext.get().getMaster().getAppMaster().getAppContext().getMasterService().getLocation();
        TConnection connection = TConnectionManager.getConnection(ps.getConf());
        MasterProtocol master = connection.getMasterService(masterLoc.getIp(), masterLoc.getPort());
        int task0Iteration = 2;
        int task1Iteration = 1;
        int task0w1Clock = 10;
        int task0w2Clock = 20;
        int task1w1Clock = 9;
        int task1w2Clock = 19;
        int w1Clock = (task0w1Clock < task1w1Clock) ? task0w1Clock : task1w1Clock;
        int w2Clock = (task0w2Clock < task1w2Clock) ? task0w2Clock : task1w2Clock;
        master.taskIteration(null, TaskIterationRequest.newBuilder().setIteration(task0Iteration).setTaskId(ProtobufUtil.convertToIdProto(task0Id)).build());
        master.taskIteration(null, TaskIterationRequest.newBuilder().setIteration(task1Iteration).setTaskId(ProtobufUtil.convertToIdProto(task1Id)).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task0Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w1Id).setClock(task0w1Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task0Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w2Id).setClock(task0w2Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task1Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w1Id).setClock(task1w1Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task1Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w2Id).setClock(task1w2Clock).build()).build());
        int writeIntervalMS = LocalClusterContext.get().getConf().getInt(AngelConf.ANGEL_AM_WRITE_STATE_INTERVAL_MS, AngelConf.DEFAULT_ANGEL_AM_WRITE_STATE_INTERVAL_MS);
        Thread.sleep(writeIntervalMS * 2);
        angelAppMaster.getAppContext().getEventHandler().handle(new InternalErrorEvent(angelAppMaster.getAppContext().getApplicationId(), "failed", true));
        Thread.sleep(10000);
        angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        assertEquals(angelAppMaster.getAppContext().getApp().getExternAppState(), AppState.RUNNING);
        LOG.info("angelAppMaster.getAppContext().getApplicationAttemptId()=" + angelAppMaster.getAppContext().getApplicationAttemptId());
        assertEquals(angelAppMaster.getAppContext().getApplicationAttemptId(), appAttempt2Id);
        PartitionKey w1Part0Key = new PartitionKey(0, w1Id, 0, 0, 1, 50000);
        PartitionKey w1Part1Key = new PartitionKey(1, w1Id, 0, 50000, 1, 100000);
        PartitionKey w2Part0Key = new PartitionKey(0, w2Id, 0, 0, 1, 50000);
        PartitionKey w2Part1Key = new PartitionKey(1, w2Id, 0, 50000, 1, 100000);
        Worker worker = LocalClusterContext.get().getWorker(worker0Attempt0Id).getWorker();
        LOG.info("worker=" + worker);
        LOG.info("worker.getTaskManager()=" + worker.getTaskManager());
        LOG.info("worker.getTaskManager().getRunningTask()=" + worker.getTaskManager().getRunningTask().size());
        TaskContext task0Context = worker.getTaskManager().getRunningTask().get(task0Id).getTaskContext();
        TaskContext task1Context = worker.getTaskManager().getRunningTask().get(task1Id).getTaskContext();
        assertEquals(task0Context.getEpoch(), task0Iteration);
        assertEquals(task1Context.getEpoch(), task1Iteration);
        LOG.info("===============worker.getPSAgent().getMatrixMetaManager().getMatrixMetas().size()=" + worker.getPSAgent().getMatrixMetaManager().getMatrixMetas().size());
        assertTrue(worker.getPSAgent().getMatrixMetaManager().exist(w1Id));
        assertTrue(worker.getPSAgent().getMatrixMetaManager().exist(w2Id));
        assertEquals(worker.getPSAgent().getMatrixMetaManager().getPss(w1Part0Key).get(0), psId);
        assertEquals(worker.getPSAgent().getMatrixMetaManager().getPss(w1Part1Key).get(0), psId);
        assertEquals(worker.getPSAgent().getMatrixMetaManager().getPss(w2Part0Key).get(0), psId);
        assertEquals(worker.getPSAgent().getMatrixMetaManager().getPss(w2Part1Key).get(0), psId);
        ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock, w2Clock);
        angelAppMaster.getAppContext().getEventHandler().handle(new InternalErrorEvent(angelAppMaster.getAppContext().getApplicationId(), "failed", true));
        Thread.sleep(10000);
        angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        assertEquals(angelAppMaster.getAppContext().getApplicationAttemptId(), appAttempt2Id);
        assertEquals(angelAppMaster.getAppContext().getApp().getExternAppState(), AppState.FAILED);
    } catch (Exception x) {
        LOG.error("run testMasterRecover failed ", x);
        throw x;
    }
}

Example 5 with ParameterServer

use of com.tencent.angel.ps.ParameterServer in project angel by Tencent.

the class InitNeighborTest method testCSR.

@Test
public void testCSR() throws Exception {
    Worker worker = LocalClusterContext.get().getWorker(workerAttempt0Id).getWorker();
    MatrixClient client = worker.getPSAgent().getMatrixClient(SPARSE_INT_MAT, 0);
    int matrixId = client.getMatrixId();
    ParameterServer ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
    Location masterLoc = LocalClusterContext.get().getMaster().getAppMaster().getAppContext().getMasterService().getLocation();
    TConnection connection = TConnectionManager.getConnection(ps.getConf());
    MasterProtocol master = connection.getMasterService(masterLoc.getIp(), masterLoc.getPort());
    // Init node neighbors
    Int2ObjectOpenHashMap<int[]> nodeIdToNeighbors = new Int2ObjectOpenHashMap<>();
    nodeIdToNeighbors.put(1, new int[] { 2, 3 });
    nodeIdToNeighbors.put(2, new int[] { 4 });
    InitNeighbor func = new InitNeighbor(new InitNeighborParam(matrixId, nodeIdToNeighbors));
    client.asyncUpdate(func).get();
    nodeIdToNeighbors.clear();
    nodeIdToNeighbors.put(1, new int[] { 4, 5, 6 });
    nodeIdToNeighbors.put(2, new int[] { 5 });
    nodeIdToNeighbors.put(4, new int[] { 5, 6 });
    func = new InitNeighbor(new InitNeighborParam(matrixId, nodeIdToNeighbors));
    client.asyncUpdate(func).get();
    nodeIdToNeighbors.clear();
    nodeIdToNeighbors.put(3, new int[] { 4, 5, 6 });
    nodeIdToNeighbors.put(5, new int[] { 6 });
    nodeIdToNeighbors.put(8, new int[] { 3, 4 });
    func = new InitNeighbor(new InitNeighborParam(matrixId, nodeIdToNeighbors));
    client.asyncUpdate(func).get();
    nodeIdToNeighbors.clear();
    client.asyncUpdate(new InitNeighborOver(new InitNeighborOverParam(matrixId))).get();
    // Sample the neighbors
    int[] nodeIds = new int[] { 1, 2, 3, 4, 5, 6, 7, 8 };
    SampleNeighborParam param = new SampleNeighborParam(matrixId, nodeIds, -1);
    Int2ObjectOpenHashMap<int[]> result = ((SampleNeighborResult) (client.get(new SampleNeighbor(param)))).getNodeIdToNeighbors();
    ObjectIterator<Entry<int[]>> iter = result.int2ObjectEntrySet().fastIterator();
    LOG.info("==============================sample neighbors result============================");
    Entry<int[]> entry;
    while (iter.hasNext()) {
        entry = iter.next();
        LOG.info("node id = " + entry.getIntKey() + ", neighbors = " + Arrays.toString(entry.getValue()));
    }
    client.checkpoint(0);
    ps.stop(-1);
    PSErrorRequest request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt0Id)).setMsg("out of memory").build();
    master.psError(null, request);
    Thread.sleep(10000);
    result = ((SampleNeighborResult) (client.get(new SampleNeighbor(param)))).getNodeIdToNeighbors();
    iter = result.int2ObjectEntrySet().fastIterator();
    LOG.info("==============================sample neighbors result============================");
    while (iter.hasNext()) {
        entry = iter.next();
        LOG.info("node id = " + entry.getIntKey() + ", neighbors = " + Arrays.toString(entry.getValue()));
    }
}