dengzhhu653 commented on code in PR #6595: URL: https://github.com/apache/hive/pull/6595#discussion_r4594742025
########## standalone-metastore/metastore-search/src/main/java/org/apache/hive/search/metastore/MetastoreEventHandler.java: ########## @@ -0,0 +1,375 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.hive.search.metastore; + +import java.io.Closeable; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collection; +import java.util.Collections; +import java.util.List; +import java.util.Objects; +import java.util.concurrent.atomic.AtomicBoolean; + +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.hive.common.DatabaseName; +import org.apache.hadoop.hive.common.TableName; +import org.apache.hadoop.hive.metastore.api.NotificationEvent; +import org.apache.hadoop.hive.metastore.api.NotificationEventRequest; +import org.apache.hadoop.hive.metastore.api.NotificationEventResponse; +import org.apache.hadoop.hive.metastore.api.Table; +import org.apache.hadoop.hive.metastore.messaging.AlterTableMessage; +import org.apache.hadoop.hive.metastore.messaging.CreateTableMessage; +import org.apache.hadoop.hive.metastore.messaging.MessageBuilder; +import org.apache.hadoop.hive.metastore.messaging.MessageDeserializer; +import org.apache.hadoop.hive.metastore.messaging.MessageFactory; +import org.apache.hive.search.exception.IndexNotHealthyException; +import org.apache.hive.search.config.IndexOptions; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** + * Polls HMS notification events, coalesces them into batches, and dispatches index mutations. + * + * <p>On repeated batch failure the handler falls back to single-event apply so one poison event + * does not block the entire batch forever. If a poison event still cannot be applied, the index + * is marked unhealthy and search is blocked until the root cause is fixed. + */ +public class MetastoreEventHandler implements Closeable { + private static final Logger LOG = LoggerFactory.getLogger(MetastoreEventHandler.class); + private final AtomicBoolean stopped = new AtomicBoolean(false); + private final List<MetastoreEventListener> listeners = Collections.synchronizedList(new ArrayList<>()); + private final IndexOptions indexConfig; + private final MetastoreClientProvider clientProvider; + private final MessageDeserializer deserializer; + private Thread eventPoller; + private long lastEventId; + + /** Consecutive batch failures at {@link #failedBatchStartId}. */ + private int consecutiveBatchFailures; + private long failedBatchStartId = -1; + + MetastoreEventHandler(Configuration configuration, MetastoreClientProvider clientProvider) { + Configuration conf = new Configuration(Objects.requireNonNull(configuration)); + this.indexConfig = new IndexOptions(conf); + this.clientProvider = Objects.requireNonNull(clientProvider); + this.deserializer = MessageFactory.getDefaultInstance(conf).getDeserializer(); + } + + public MetastoreEventHandler addListeners(MetastoreEventListener... listeners) { + return addListeners(Arrays.asList(listeners)); + } + + public MetastoreEventHandler addListeners(Collection<MetastoreEventListener> listeners) { + this.listeners.addAll(listeners); + return this; + } + + public MetastoreEventHandler start(long nid) throws Exception { + this.lastEventId = nid; + long sleepInterval = indexConfig.getPollNotificationInterval(); + int pollMax = indexConfig.getEventPollMax(); + long unhealthyBackoffMs = indexConfig.getEventUnhealthyBackoffMs(); + this.eventPoller = new Thread(() -> { + while (!stopped.get()) { + try { + int processed = getNextMetastoreEvents(pollMax); + if (processed <= 0) { + Thread.sleep(sleepInterval); + } + } catch (InterruptedException e) { + Thread.currentThread().interrupt(); + break; + } catch (IndexNotHealthyException e) { + LOG.error("Incremental index update halted: {}", e.getMessage(), e); + notifyListenersStatus(false, e); + try { + Thread.sleep(unhealthyBackoffMs); + } catch (InterruptedException ex) { + Thread.currentThread().interrupt(); + break; + } + } + } + }); + this.eventPoller.setName("Metastore-Event-Poller"); + this.eventPoller.start(); + return this; + } + + @Override + public void close() { + stopped.set(true); + try { + if (eventPoller != null) { + eventPoller.interrupt(); + eventPoller.join(); + } + } catch (InterruptedException ignored) { + // ignore + } + } + + public int getNextMetastoreEvents(int eventCount) throws IndexNotHealthyException { + try { + NotificationEventRequest request = new NotificationEventRequest(lastEventId); + request.setMaxEvents(eventCount); + final NotificationEventResponse resp; + try (MetastoreClientProvider.MetastoreClientBundle clientBundle = + clientProvider.getMetastoreClient(indexConfig.configuration())) { + resp = clientBundle.client().getNextNotification(request, true, null); + if (resp == null || resp.getEvents() == null || resp.getEvents().isEmpty()) { + LOG.debug("No event found since the last event id: {}", lastEventId); + return 0; + } + } + + List<NotificationEvent> events = resp.getEvents(); + long batchStartId = events.get(0).getEventId(); + MetastoreEventListener.IndexTask task; + try { + task = buildTask(events); + } catch (Exception parseError) { + LOG.warn( + "Failed to build notification batch starting at event {}; falling back to single-event apply", + batchStartId, + parseError); + int applied = applyEventsIndividually(events); + if (applied > 0) { + notifyListenersStatus(true); + resetBatchFailureState(); + return applied; + } + backoffAfterFailure(); + return 0; + } + + try { + notifyListeners(task); + lastEventId = task.lastEventId; + notifyListenersStatus(true); + resetBatchFailureState(); + return events.size(); + } catch (Exception batchError) { + recordBatchFailure(batchStartId, task, batchError); + if (consecutiveBatchFailures >= indexConfig.getEventBatchMaxFailures()) { + LOG.warn( + "Batch apply failed {} time(s) at event {}; falling back to single-event apply", + consecutiveBatchFailures, + failedBatchStartId); + int applied = applyEventsIndividually(events); + if (applied > 0) { + notifyListenersStatus(true); + resetBatchFailureState(); + return applied; + } + } + backoffAfterFailure(); + return 0; + } + } catch (IndexNotHealthyException e) { + throw e; + } catch (Exception e) { + LOG.warn("Failed to fetch or parse notification events after lastEventId={}", lastEventId, e); + backoffAfterFailure(); + } + return 0; + } + + private int applyEventsIndividually(List<NotificationEvent> events) + throws IndexNotHealthyException { + int applied = 0; + for (NotificationEvent event : events) { + if (lastEventId >= event.getEventId()) { + continue; + } + MetastoreEventListener.IndexTask task = + new MetastoreEventListener.IndexTask(); + task.firstEventId = event.getEventId(); + try { + mergeEvent(event, task); + notifyListeners(task); + lastEventId = event.getEventId(); + applied++; + } catch (Exception e) { + long poisonEventId = event.getEventId(); + String poisonEventType = event.getEventType(); + if (indexConfig.isEventSkipPoison()) { + LOG.error( + "Skipping poison notification event {} (type={}); index may be stale for this object", + poisonEventId, + poisonEventType, + e); + lastEventId = event.getEventId(); + applied++; + } else { + LOG.error( + "Stuck on notification event {} (type={}); committed lastEventId={}", + poisonEventId, + poisonEventType, + lastEventId, + e); + String progress = applied > 0 + ? applied + " prior event(s) in batch were applied; " + : ""; + throw new IndexNotHealthyException( + "Cannot apply notification event " + + poisonEventId + + " (type=" + + poisonEventType + + "); " + + progress + + "committed lastEventId=" + + lastEventId + + ". Fix root cause, set " + + IndexOptions.EVENT_SKIP_POISON + + "=true to skip, or rebuild the index.", + e); + } + } + } + if (applied > 0) { + LOG.info("Applied {} notification event(s) individually; lastEventId={}", applied, lastEventId); + } + return applied; + } + + private MetastoreEventListener.IndexTask buildTask(List<NotificationEvent> events) + throws Exception { + MetastoreEventListener.IndexTask task = new MetastoreEventListener.IndexTask(); + for (int i = 0; i < events.size(); i++) { + NotificationEvent event = events.get(i); + if (lastEventId >= event.getEventId()) { + // this should not happen + throw new IllegalStateException( + "Out-of-order metastore notification event: lastEventId=" + lastEventId + + ", eventId=" + event.getEventId()); + } + if (i == 0) { + task.firstEventId = event.getEventId(); + } + mergeEvent(event, task); + } + return task; + } + + private void notifyListeners(MetastoreEventListener.IndexTask task) throws Exception { + for (MetastoreEventListener listener : listeners) { + listener.notifyIndexTask(task); + } + } + + private void notifyListenersStatus(boolean healthy, IndexNotHealthyException... errors) { + for (MetastoreEventListener listener : listeners) { + listener.notifyIndexState(healthy, errors); + } + } + + private void recordBatchFailure( + long batchStartId, MetastoreEventListener.IndexTask task, Exception e) { + if (failedBatchStartId != batchStartId) { + failedBatchStartId = batchStartId; + consecutiveBatchFailures = 1; + } else { + consecutiveBatchFailures++; + } + LOG.warn( + "Failed to apply notification batch (attempt {}/{}, committed lastEventId={}, " + + "batch event range {}-{}, upserts={}, tableDrops={}, dbDrops={})", + consecutiveBatchFailures, + indexConfig.getEventBatchMaxFailures(), + lastEventId, + task.firstEventId, + task.lastEventId, + task.tablesToAdd.size(), + task.tablesToDrop.size(), + task.databasesToDrop.size(), + e); + } + + private void resetBatchFailureState() { + consecutiveBatchFailures = 0; + failedBatchStartId = -1; + } + + private void backoffAfterFailure() { + long backoffMs = indexConfig.getEventFailureBackoffMs(); + if (backoffMs <= 0) { + return; + } + try { + Thread.sleep(backoffMs); + } catch (InterruptedException e) { + Thread.currentThread().interrupt(); + } + } + + private void mergeEvent( + NotificationEvent event, + MetastoreEventListener.IndexTask task) + throws Exception { + String message = event.getMessage(); + switch (event.getEventType()) { + case MessageBuilder.CREATE_TABLE_EVENT -> { + CreateTableMessage createTableMessage = deserializer.getCreateTableMessage(message); + Table table = createTableMessage.getTableObj(); + TableName tableName = new TableName(table.getCatName(), table.getDbName(), table.getTableName()); + task.tablesToAdd.put(tableName, table); + } + case MessageBuilder.DROP_TABLE_EVENT -> { Review Comment: > its CREATE and DROP events replay in one batch and cancel out Very good catch!!! > could DROP always add to tablesToDrop No, in case if the table is re-created after the drop We need special handling of the notifications within bootstrap overlaps. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
