use of com.linkedin.thirdeye.client.pinot.PinotQuery in project pinot by linkedin.
the class ThirdEyeCacheRegistry method initCaches.
private static void initCaches(ThirdEyeConfiguration config) {
ThirdEyeCacheRegistry cacheRegistry = ThirdEyeCacheRegistry.getInstance();
RemovalListener<PinotQuery, ResultSetGroup> listener = new RemovalListener<PinotQuery, ResultSetGroup>() {
@Override
public void onRemoval(RemovalNotification<PinotQuery, ResultSetGroup> notification) {
LOGGER.info("Expired {}", notification.getKey().getPql());
}
};
// ResultSetGroup Cache. The size of this cache is limited by the total number of buckets in all ResultSetGroup.
// We estimate that 1 bucket (including overhead) consumes 1KB and this cache is allowed to use up to 50% of max
// heap space.
long maxBucketNumber = getApproximateMaxBucketNumber(DEFAULT_HEAP_PERCENTAGE_FOR_RESULTSETGROUP_CACHE);
LoadingCache<PinotQuery, ResultSetGroup> resultSetGroupCache = CacheBuilder.newBuilder().removalListener(listener).expireAfterAccess(1, TimeUnit.HOURS).maximumWeight(maxBucketNumber).weigher((pinotQuery, resultSetGroup) -> {
int resultSetCount = resultSetGroup.getResultSetCount();
int weight = 0;
for (int idx = 0; idx < resultSetCount; ++idx) {
com.linkedin.pinot.client.ResultSet resultSet = resultSetGroup.getResultSet(idx);
weight += (resultSet.getColumnCount() * resultSet.getRowCount());
}
return weight;
}).build(new ResultSetGroupCacheLoader(pinotThirdeyeClientConfig));
cacheRegistry.registerResultSetGroupCache(resultSetGroupCache);
LOGGER.info("Max bucket number for ResultSetGroup cache is set to {}", maxBucketNumber);
// CollectionMaxDataTime Cache
LoadingCache<String, Long> collectionMaxDataTimeCache = CacheBuilder.newBuilder().refreshAfterWrite(5, TimeUnit.MINUTES).build(new CollectionMaxDataTimeCacheLoader(resultSetGroupCache, datasetConfigDAO));
cacheRegistry.registerCollectionMaxDataTimeCache(collectionMaxDataTimeCache);
// Query Cache
QueryCache queryCache = new QueryCache(thirdEyeClient, Executors.newFixedThreadPool(10));
cacheRegistry.registerQueryCache(queryCache);
// Dimension Filter cache
LoadingCache<String, String> dimensionFiltersCache = CacheBuilder.newBuilder().build(new DimensionFiltersCacheLoader(cacheRegistry.getQueryCache()));
cacheRegistry.registerDimensionFiltersCache(dimensionFiltersCache);
// Dashboards cache
LoadingCache<String, String> dashboardsCache = CacheBuilder.newBuilder().build(new DashboardsCacheLoader(dashboardConfigDAO));
cacheRegistry.registerDashboardsCache(dashboardsCache);
// Collections cache
CollectionsCache collectionsCache = new CollectionsCache(datasetConfigDAO, config);
cacheRegistry.registerCollectionsCache(collectionsCache);
// DatasetConfig cache
LoadingCache<String, DatasetConfigDTO> datasetConfigCache = CacheBuilder.newBuilder().build(new DatasetConfigCacheLoader(datasetConfigDAO));
cacheRegistry.registerDatasetConfigCache(datasetConfigCache);
// MetricConfig cache
LoadingCache<MetricDataset, MetricConfigDTO> metricConfigCache = CacheBuilder.newBuilder().build(new MetricConfigCacheLoader(metricConfigDAO));
cacheRegistry.registerMetricConfigCache(metricConfigCache);
// DashboardConfigs cache
LoadingCache<String, List<DashboardConfigDTO>> dashboardConfigsCache = CacheBuilder.newBuilder().build(new DashboardConfigCacheLoader(dashboardConfigDAO));
cacheRegistry.registerDashboardConfigsCache(dashboardConfigsCache);
}
use of com.linkedin.thirdeye.client.pinot.PinotQuery in project pinot by linkedin.
the class AutoLoadPinotMetricsService method fetchMetricAsADimensionMetrics.
private List<String> fetchMetricAsADimensionMetrics(String dataset, String metricNamesColumn) {
List<String> distinctMetricNames = new ArrayList<>();
ThirdEyeCacheRegistry CACHE_REGISTRY = ThirdEyeCacheRegistry.getInstance();
String sql = String.format("select count(*) from %s group by %s top 10", dataset, metricNamesColumn);
try {
ResultSetGroup result = CACHE_REGISTRY.getResultSetGroupCache().get(new PinotQuery(sql, dataset));
ResultSet resultSet = result.getResultSet(0);
int rowCount = resultSet.getRowCount();
for (int i = 0; i < rowCount; i++) {
String dimensionName = resultSet.getGroupKeyString(i, 0);
distinctMetricNames.add(dimensionName);
}
LOG.info("Distinct Metrics {}", distinctMetricNames);
} catch (Exception e) {
LOG.error("Exception in fetching metrics from pinot", e);
}
return distinctMetricNames;
}
use of com.linkedin.thirdeye.client.pinot.PinotQuery in project pinot by linkedin.
the class AnomalyApplicationEndToEndTest method setup.
private void setup() throws Exception {
// Mock query cache
ThirdEyeClient mockThirdeyeClient = Mockito.mock(ThirdEyeClient.class);
Mockito.when(mockThirdeyeClient.execute(Matchers.any(ThirdEyeRequest.class))).thenAnswer(new Answer<ThirdEyeResponse>() {
@Override
public ThirdEyeResponse answer(InvocationOnMock invocation) throws Throwable {
Object[] args = invocation.getArguments();
ThirdEyeRequest request = (ThirdEyeRequest) args[0];
ThirdEyeResponse response = getMockResponse(request);
return response;
}
});
QueryCache mockQueryCache = new QueryCache(mockThirdeyeClient, Executors.newFixedThreadPool(10));
cacheRegistry.registerQueryCache(mockQueryCache);
MetricConfigDTO metricConfig = getTestMetricConfig(collection, metric, 1L);
// create metric config in cache
LoadingCache<MetricDataset, MetricConfigDTO> mockMetricConfigCache = Mockito.mock(LoadingCache.class);
Mockito.when(mockMetricConfigCache.get(new MetricDataset(metric, collection))).thenReturn(metricConfig);
cacheRegistry.registerMetricConfigCache(mockMetricConfigCache);
// create dataset config in cache
LoadingCache<String, DatasetConfigDTO> mockDatasetConfigCache = Mockito.mock(LoadingCache.class);
Mockito.when(mockDatasetConfigCache.get(collection)).thenReturn(getTestDatasetConfig(collection));
cacheRegistry.registerDatasetConfigCache(mockDatasetConfigCache);
ResultSet mockResultSet = Mockito.mock(ResultSet.class);
Mockito.when(mockResultSet.getRowCount()).thenReturn(0);
ResultSetGroup mockResultSetGroup = Mockito.mock(ResultSetGroup.class);
Mockito.when(mockResultSetGroup.getResultSet(0)).thenReturn(mockResultSet);
LoadingCache<PinotQuery, ResultSetGroup> mockResultSetGroupCache = Mockito.mock(LoadingCache.class);
Mockito.when(mockResultSetGroupCache.get(Matchers.any(PinotQuery.class))).thenAnswer(new Answer<ResultSetGroup>() {
@Override
public ResultSetGroup answer(InvocationOnMock invocation) throws Throwable {
return mockResultSetGroup;
}
});
cacheRegistry.registerResultSetGroupCache(mockResultSetGroupCache);
// Application config
thirdeyeAnomalyConfig = new ThirdEyeAnomalyConfiguration();
thirdeyeAnomalyConfig.setId(id);
thirdeyeAnomalyConfig.setDashboardHost(dashboardHost);
MonitorConfiguration monitorConfiguration = new MonitorConfiguration();
monitorConfiguration.setMonitorFrequency(new TimeGranularity(30, TimeUnit.SECONDS));
thirdeyeAnomalyConfig.setMonitorConfiguration(monitorConfiguration);
thirdeyeAnomalyConfig.setRootDir(System.getProperty("dw.rootDir", "NOT_SET(dw.rootDir)"));
// create test anomaly function
functionId = anomalyFunctionDAO.save(getTestFunctionSpec(metric, collection));
// create test email configuration
emailConfigurationDAO.save(getTestEmailConfiguration(metric, collection));
// create test alert configuration
alertConfigDAO.save(getTestAlertConfiguration("test alert v2"));
// create test dataset config
datasetConfigDAO.save(getTestDatasetConfig(collection));
// setup function factory for worker and merger
InputStream factoryStream = AnomalyApplicationEndToEndTest.class.getResourceAsStream(functionPropertiesFile);
anomalyFunctionFactory = new AnomalyFunctionFactory(factoryStream);
// setup alertfilter factory for worker
InputStream alertFilterStream = AnomalyApplicationEndToEndTest.class.getResourceAsStream(alertFilterPropertiesFile);
alertFilterFactory = new AlertFilterFactory(alertFilterStream);
}
use of com.linkedin.thirdeye.client.pinot.PinotQuery in project pinot by linkedin.
the class CollectionMaxDataTimeCacheLoader method load.
@Override
public Long load(String collection) throws Exception {
LOGGER.info("Loading maxDataTime cache {}", collection);
long maxTime = 0;
try {
DatasetConfigDTO datasetConfig = datasetConfigDAO.findByDataset(collection);
// By default, query only offline, unless dataset has been marked as realtime
String tableName = ThirdEyeUtils.computeTableName(collection);
TimeSpec timeSpec = ThirdEyeUtils.getTimeSpecFromDatasetConfig(datasetConfig);
long prevMaxDataTime = getPrevMaxDataTime(collection, timeSpec);
String maxTimePql = String.format(COLLECTION_MAX_TIME_QUERY_TEMPLATE, timeSpec.getColumnName(), tableName, timeSpec.getColumnName(), prevMaxDataTime);
PinotQuery maxTimePinotQuery = new PinotQuery(maxTimePql, tableName);
resultSetGroupCache.refresh(maxTimePinotQuery);
ResultSetGroup resultSetGroup = resultSetGroupCache.get(maxTimePinotQuery);
if (resultSetGroup.getResultSetCount() == 0 || resultSetGroup.getResultSet(0).getRowCount() == 0) {
LOGGER.info("resultSetGroup is Empty for collection {} is {}", tableName, resultSetGroup);
this.collectionToPrevMaxDataTimeMap.remove(collection);
} else {
long endTime = new Double(resultSetGroup.getResultSet(0).getDouble(0)).longValue();
this.collectionToPrevMaxDataTimeMap.put(collection, endTime);
// endTime + 1 to make sure we cover the time range of that time value.
String timeFormat = timeSpec.getFormat();
if (StringUtils.isBlank(timeFormat) || TimeSpec.SINCE_EPOCH_FORMAT.equals(timeFormat)) {
maxTime = timeSpec.getDataGranularity().toMillis(endTime + 1) - 1;
} else {
DateTimeFormatter inputDataDateTimeFormatter = DateTimeFormat.forPattern(timeFormat).withZone(Utils.getDataTimeZone(collection));
maxTime = DateTime.parse(String.valueOf(endTime), inputDataDateTimeFormatter).getMillis();
}
}
} catch (Exception e) {
LOGGER.warn("Exception getting maxTime from collection: {}", collection, e);
this.collectionToPrevMaxDataTimeMap.remove(collection);
}
if (maxTime <= 0) {
maxTime = System.currentTimeMillis();
}
return maxTime;
}
use of com.linkedin.thirdeye.client.pinot.PinotQuery in project pinot by linkedin.
the class DataCompletenessTaskUtils method getBucketNameToCountStarMap.
private static Map<String, Long> getBucketNameToCountStarMap(String dataset, TimeSpec timeSpec, ListMultimap<String, Long> bucketNameToTimeValues) {
Map<String, Long> bucketNameToCountStar = new HashMap<>();
// generate request
StringBuilder sb = new StringBuilder();
String delimiter = "";
for (Long timeValue : bucketNameToTimeValues.values()) {
sb.append(delimiter);
delimiter = " OR ";
sb.append(String.format("%s='%s'", timeSpec.getColumnName(), timeValue));
}
long top = bucketNameToTimeValues.values().size();
String pql = String.format("select count(*) from %s where %s group by %s top %s", dataset, sb.toString(), timeSpec.getColumnName(), top);
Map<Long, Long> timeValueToCount = new HashMap<>();
try {
ResultSetGroup resultSetGroup = CACHE_REGISTRY.getResultSetGroupCache().get(new PinotQuery(pql, dataset));
if (resultSetGroup == null || resultSetGroup.getResultSetCount() <= 0) {
return bucketNameToCountStar;
}
ResultSet resultSet = resultSetGroup.getResultSet(0);
for (int i = 0; i < resultSet.getRowCount(); i++) {
Long timeValue = Long.valueOf(resultSet.getGroupKeyString(i, 0));
Long count = resultSet.getLong(i, 0);
timeValueToCount.put(timeValue, count);
}
} catch (ExecutionException e) {
LOG.error("Exception in getting count *. PQL:{}", pql, e);
}
// parse response to get counts
for (String bucketName : bucketNameToTimeValues.keySet()) {
List<Long> timeValues = bucketNameToTimeValues.get(bucketName);
Long sumOfCountForBucket = 0L;
for (Long timeValue : timeValues) {
sumOfCountForBucket = sumOfCountForBucket + timeValueToCount.getOrDefault(timeValue, 0L);
}
bucketNameToCountStar.put(bucketName, sumOfCountForBucket);
}
return bucketNameToCountStar;
}
Aggregations