Examples with Symptom com.microsoft.dhalion.detector.Symptom used on opensource projects

Example 11 with Symptom

use of com.microsoft.dhalion.detector.Symptom in project incubator-heron by apache.

the class UnderProvisioningDiagnoserTest method diagnosisFailsNotSimilarProcessingRates.

@Test
public void diagnosisFailsNotSimilarProcessingRates() {
    List<Symptom> symptoms = TestUtils.createBpSymptomList(123, 0, 0);
    symptoms.add(TestUtils.createExeCountSymptom(100, 500, 500));
    Diagnosis result = new UnderProvisioningDiagnoser().diagnose(symptoms);
    assertNull(result);
}

Example 12 with Symptom

use of com.microsoft.dhalion.detector.Symptom in project incubator-heron by apache.

the class GrowingWaitQueueDetector method detect.

/**
 * Detects all components unable to keep up with input load, hence having a growing pending buffer
 * or wait queue
 *
 * @return A collection of all components executing slower than input rate.
 */
@Override
public List<Symptom> detect() {
    ArrayList<Symptom> result = new ArrayList<>();
    Map<String, ComponentMetrics> bufferSizes = pendingBufferSensor.get();
    for (ComponentMetrics compMetrics : bufferSizes.values()) {
        ComponentMetricsHelper compStats = new ComponentMetricsHelper(compMetrics);
        compStats.computeBufferSizeTrend();
        if (compStats.getMaxBufferChangeRate() > rateLimit) {
            LOG.info(String.format("Detected growing wait queues for %s, max rate %f", compMetrics.getName(), compStats.getMaxBufferChangeRate()));
            result.add(new Symptom(SYMPTOM_GROWING_WAIT_Q.text(), compMetrics));
        }
    }
    return result;
}

Example 13 with Symptom

use of com.microsoft.dhalion.detector.Symptom in project incubator-heron by apache.

the class SlowInstanceDiagnoser method diagnose.

@Override
public Diagnosis diagnose(List<Symptom> symptoms) {
    List<Symptom> bpSymptoms = getBackPressureSymptoms(symptoms);
    Map<String, ComponentMetrics> processingRateSkewComponents = getProcessingRateSkewComponents(symptoms);
    Map<String, ComponentMetrics> waitQDisparityComponents = getWaitQDisparityComponents(symptoms);
    if (bpSymptoms.isEmpty() || waitQDisparityComponents.isEmpty() || !processingRateSkewComponents.isEmpty()) {
        // execution count, no action is needed
        return null;
    } else if (bpSymptoms.size() > 1) {
        // TODO handle cases where multiple detectors create back pressure symptom
        throw new IllegalStateException("Multiple back-pressure symptoms case");
    }
    ComponentMetrics bpMetrics = bpSymptoms.iterator().next().getComponent();
    // verify wait Q disparity and back pressure for the same component exists
    ComponentMetrics pendingBufferMetrics = waitQDisparityComponents.get(bpMetrics.getName());
    if (pendingBufferMetrics == null) {
        // no wait Q disparity for the component with back pressure. There is no slow instance
        return null;
    }
    ComponentMetrics mergedData = ComponentMetrics.merge(bpMetrics, pendingBufferMetrics);
    ComponentMetricsHelper compStats = new ComponentMetricsHelper(mergedData);
    compStats.computeBpStats();
    MetricsStats bufferStats = compStats.computeMinMaxStats(METRIC_BUFFER_SIZE);
    Symptom resultSymptom = null;
    for (InstanceMetrics boltMetrics : compStats.getBoltsWithBackpressure()) {
        double bufferSize = boltMetrics.getMetricValueSum(METRIC_BUFFER_SIZE.text());
        double bpValue = boltMetrics.getMetricValueSum(METRIC_BACK_PRESSURE.text());
        if (bufferStats.getMetricMax() < bufferSize * 2) {
            LOG.info(String.format("SLOW: %s back-pressure(%s) and high buffer size: %s " + "and similar processing rates", boltMetrics.getName(), bpValue, bufferSize));
            resultSymptom = new Symptom(SYMPTOM_SLOW_INSTANCE.text(), mergedData);
        }
    }
    return resultSymptom != null ? new Diagnosis(DIAGNOSIS_SLOW_INSTANCE.text(), resultSymptom) : null;
}

Example 14 with Symptom

use of com.microsoft.dhalion.detector.Symptom in project incubator-heron by apache.

the class SkewDetector method detect.

/**
 * Detects components experiencing data skew, instances with vastly different execute counts.
 *
 * @return A collection of affected components
 */
@Override
public List<Symptom> detect() {
    ArrayList<Symptom> result = new ArrayList<>();
    Map<String, ComponentMetrics> metrics = sensor.get();
    for (ComponentMetrics compMetrics : metrics.values()) {
        ComponentMetricsHelper compStats = new ComponentMetricsHelper(compMetrics);
        MetricsStats stats = compStats.computeMinMaxStats(sensor.getMetricName());
        if (stats.getMetricMax() > skewRatio * stats.getMetricMin()) {
            LOG.info(String.format("Detected skew for %s, min = %f, max = %f", compMetrics.getName(), stats.getMetricMin(), stats.getMetricMax()));
            result.add(new Symptom(symptomName.text(), compMetrics));
        }
    }
    return result;
}

Example 15 with Symptom

use of com.microsoft.dhalion.detector.Symptom in project incubator-heron by apache.

the class RestartContainerResolver method resolve.

@Override
public List<Action> resolve(List<Diagnosis> diagnosis) {
    List<Action> actions = new ArrayList<>();
    for (Diagnosis diagnoses : diagnosis) {
        Symptom bpSymptom = diagnoses.getSymptoms().get(SYMPTOM_SLOW_INSTANCE.text());
        if (bpSymptom == null || bpSymptom.getComponents().isEmpty()) {
            // nothing to fix as there is no back pressure
            continue;
        }
        if (bpSymptom.getComponents().size() > 1) {
            throw new UnsupportedOperationException("Multiple components with back pressure symptom");
        }
        // want to know which stmgr has backpressure
        String stmgrId = null;
        for (InstanceMetrics im : bpSymptom.getComponent().getMetrics().values()) {
            if (im.hasMetricAboveLimit(METRIC_BACK_PRESSURE.text(), noiseFilterMillis)) {
                String instanceId = im.getName();
                int fromIndex = instanceId.indexOf('_') + 1;
                int toIndex = instanceId.indexOf('_', fromIndex);
                stmgrId = instanceId.substring(fromIndex, toIndex);
                break;
            }
        }
        LOG.info("Restarting container: " + stmgrId);
        boolean b = schedulerClient.restartTopology(RestartTopologyRequest.newBuilder().setContainerIndex(Integer.valueOf(stmgrId)).setTopologyName(topologyName).build());
        LOG.info("Restarted container result: " + b);
        ContainerRestart action = new ContainerRestart();
        LOG.info("Broadcasting container restart event");
        eventManager.onEvent(action);
        actions.add(action);
        return actions;
    }
    return actions;
}