stevenzwu commented on code in PR #10308: URL: https://github.com/apache/iceberg/pull/10308#discussion_r1608678483
########## flink/v1.19/flink/src/main/java/org/apache/iceberg/flink/maintenance/operator/MonitorSource.java: ########## @@ -0,0 +1,200 @@ +/* + * 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.iceberg.flink.maintenance.operator; + +import java.io.IOException; +import java.util.Iterator; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.api.connector.source.Boundedness; +import org.apache.flink.api.connector.source.SourceReader; +import org.apache.flink.api.connector.source.SourceReaderContext; +import org.apache.flink.api.connector.source.util.ratelimit.RateLimitedSourceReader; +import org.apache.flink.api.connector.source.util.ratelimit.RateLimiter; +import org.apache.flink.api.connector.source.util.ratelimit.RateLimiterStrategy; +import org.apache.flink.core.io.SimpleVersionedSerializer; +import org.apache.flink.core.memory.DataInputDeserializer; +import org.apache.flink.core.memory.DataOutputSerializer; +import org.apache.iceberg.Snapshot; +import org.apache.iceberg.Table; +import org.apache.iceberg.flink.TableLoader; +import org.apache.iceberg.relocated.com.google.common.annotations.VisibleForTesting; +import org.apache.iceberg.relocated.com.google.common.base.MoreObjects; +import org.apache.iceberg.relocated.com.google.common.base.Preconditions; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** Monitors an Iceberg table for changes */ +public class MonitorSource extends SingleThreadedIteratorSource<TableChange> { + private static final Logger LOG = LoggerFactory.getLogger(MonitorSource.class); + + private final TableLoader tableLoader; + private final RateLimiterStrategy rateLimiterStrategy; + private final long maxReadBack; + + /** + * Creates a {@link org.apache.flink.api.connector.source.Source} which monitors an Iceberg table + * for changes. + * + * @param tableLoader used for accessing the table + * @param rateLimiterStrategy limits the frequency the table is checked + * @param maxReadBack sets the number of snapshots read before stopping change collection + */ + public MonitorSource( + TableLoader tableLoader, RateLimiterStrategy rateLimiterStrategy, long maxReadBack) { + Preconditions.checkNotNull(tableLoader, "Table loader should no be null"); + Preconditions.checkNotNull(rateLimiterStrategy, "Rate limiter strategy should no be null"); + Preconditions.checkArgument(maxReadBack > 0, "Need to read at least 1 snapshot to work"); + + this.tableLoader = tableLoader; + this.rateLimiterStrategy = rateLimiterStrategy; + this.maxReadBack = maxReadBack; + } + + @Override + public Boundedness getBoundedness() { + return Boundedness.CONTINUOUS_UNBOUNDED; + } + + @Override + public TypeInformation<TableChange> getProducedType() { + return TypeInformation.of(TableChange.class); + } + + @Override + Iterator<TableChange> createIterator() { + return new SchedulerEventIterator(tableLoader, null, maxReadBack); + } + + @Override + SimpleVersionedSerializer<Iterator<TableChange>> getIteratorSerializer() { + return new SchedulerEventIteratorSerializer(tableLoader, maxReadBack); + } + + @Override + public SourceReader<TableChange, GlobalSplit<TableChange>> createReader( + SourceReaderContext readerContext) throws Exception { + RateLimiter rateLimiter = rateLimiterStrategy.createRateLimiter(1); + return new RateLimitedSourceReader<>(super.createReader(readerContext), rateLimiter); + } + + /** The Iterator which returns the latest changes on an Iceberg table. */ + @VisibleForTesting + static class SchedulerEventIterator implements Iterator<TableChange> { + private Long lastSnapshotId; + private final long maxReadBack; + private final Table table; + + SchedulerEventIterator(TableLoader tableLoader, Long lastSnapshotId, long maxReadBack) { + this.lastSnapshotId = lastSnapshotId; + this.maxReadBack = maxReadBack; + tableLoader.open(); + this.table = tableLoader.loadTable(); + } + + @Override + public boolean hasNext() { + return true; + } + + @Override + public TableChange next() { + try { + table.refresh(); + Snapshot currentSnapshot = table.currentSnapshot(); + Long current = currentSnapshot != null ? currentSnapshot.snapshotId() : null; + Long checking = current; + TableChange event = new TableChange(0, 0, 0L, 0L, 0); + long readBack = 0; + while (checking != null && !checking.equals(lastSnapshotId) && ++readBack <= maxReadBack) { Review Comment: > While the Maintenance Tasks are triggered a bit later, this doesn't effect correctness, as the Tasks themselves read the current state of the table, and the collected TableChange is only used as a trigger. this is true if we assume the scheduler would read the current state of the table for scheduling decision. Let's use small files compaction with table partitioned by ingestion time as an example. New commits typically touch 1 or 2 partitions. there is no need to scan the whole table metadata (which could be large) to determine the small files. compaction task can just check the new files for compaction candidates and possibly check the all data files from the partitions that the delta change/commit touched. that could reduce the scope of checking from whole table to small number of partitions. Often, it is wasteful to check the old partitions over and over again for compaction candidates of small files. What I am trying to say is that it can be beneficial to have incremental check and compaction (from last compacted snapshot to a new snapshot). There are certainly some tasks requiring whole table check (like orphan files cleanup). -- 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: issues-unsubscr...@iceberg.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@iceberg.apache.org For additional commands, e-mail: issues-h...@iceberg.apache.org
