aokolnychyi opened a new issue #104: ManifestReader is not properly closed in BaseTableScan URL: https://github.com/apache/incubator-iceberg/issues/104 It seems `ManifestReader` is not closed properly in some cases. `ManifestReader` is `CloseableGroup` and we add `AvroIterable` to the list of closeables. However, `ManifestReader$close` is not called. One way to reproduce this is to run `TestParquetWrite`. ``` [Finalizer] WARN com.netflix.iceberg.hadoop.HadoopStreams - Unclosed input stream created by: com.netflix.iceberg.hadoop.HadoopStreams$HadoopSeekableInputStream.<init>(HadoopStreams.java:78) com.netflix.iceberg.hadoop.HadoopStreams.wrap(HadoopStreams.java:53) com.netflix.iceberg.hadoop.HadoopInputFile.newStream(HadoopInputFile.java:125) com.netflix.iceberg.avro.AvroIterable.newFileReader(AvroIterable.java:95) com.netflix.iceberg.avro.AvroIterable.iterator(AvroIterable.java:77) com.google.common.collect.Iterables$6.iterator(Iterables.java:585) com.google.common.collect.Iterables$8.iterator(Iterables.java:709) com.netflix.iceberg.ManifestReader.iterator(ManifestReader.java:225) com.netflix.iceberg.FilteredManifest.iterator(FilteredManifest.java:124) com.google.common.collect.Iterables$8.iterator(Iterables.java:709) com.google.common.collect.Iterables$3.transform(Iterables.java:508) com.google.common.collect.Iterables$3.transform(Iterables.java:505) com.google.common.collect.TransformedIterator.next(TransformedIterator.java:48) com.google.common.collect.Iterators$5.hasNext(Iterators.java:543) com.netflix.iceberg.util.BinPacking$PackingIterator.hasNext(BinPacking.java:91) com.netflix.iceberg.io.CloseableIterable$4$1.hasNext(CloseableIterable.java:92) com.google.common.collect.Iterators.addAll(Iterators.java:356) com.google.common.collect.Lists.newArrayList(Lists.java:147) com.google.common.collect.Lists.newArrayList(Lists.java:129) com.netflix.iceberg.spark.source.Reader.tasks(Reader.java:209) com.netflix.iceberg.spark.source.Reader.planInputPartitions(Reader.java:135) org.apache.spark.sql.execution.datasources.v2.DataSourceV2ScanExec.partitions$lzycompute(DataSourceV2ScanExec.scala:76) ... ``` My suspicion is that we have a problem in `BaseTableScan$planFiles`. Specifically, this piece: ``` ConcurrentLinkedQueue<Closeable> toClose = new ConcurrentLinkedQueue<>(); Iterable<Iterable<FileScanTask>> readers = Iterables.transform( matchingManifests, manifest -> { ManifestReader reader = ManifestReader.read(ops.io().newInputFile(manifest.path())); toClose.add(reader); String schemaString = SchemaParser.toJson(reader.spec().schema()); String specString = PartitionSpecParser.toJson(reader.spec()); ResidualEvaluator residuals = new ResidualEvaluator(reader.spec(), rowFilter); return Iterables.transform( reader.filterRows(rowFilter).select(SNAPSHOT_COLUMNS), file -> new BaseFileScanTask(file, schemaString, specString, residuals) ); }); if (PLAN_SCANS_WITH_WORKER_POOL && snapshot.manifests().size() > 1) { return CloseableIterable.combine( new ParallelIterable<>(readers, getWorkerPool()), toClose); } else { return CloseableIterable.combine(Iterables.concat(readers), toClose); } ``` The closure that adds `ManifestReader reader` to `toClose` is lazy. Hence, `toClose` is empty when we return `CloseableIterable.combine(Iterables.concat(readers), toClose)`.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
