From 5cf7b0280f8151c1461300a7a61a80c436dff0a7 Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Mon, 2 Mar 2026 21:06:57 +1100 Subject: [PATCH 1/7] Support per-partition buckets --- docs/docs/flink/procedures.md | 2 +- docs/docs/maintenance/rescale-bucket.md | 26 +- .../primary-key-table/data-distribution.md | 5 +- .../operation/AbstractFileStoreWrite.java | 36 ++- .../privilege/PrivilegedFileStoreTable.java | 8 + .../paimon/table/AbstractFileStoreTable.java | 8 +- .../table/AppendOnlyFileStoreTable.java | 9 +- .../paimon/table/DelegatedFileStoreTable.java | 6 + .../apache/paimon/table/FileStoreTable.java | 7 + .../table/PrimaryKeyFileStoreTable.java | 10 +- .../paimon/table/RescaleFileStoreTable.java | 92 +++++++ .../sink/FixedBucketRowKeyExtractor.java | 37 ++- .../table/sink/FixedBucketWriteSelector.java | 7 +- .../table/sink/PartitionBucketMapping.java | 130 ++++++++++ .../paimon/table/sink/RowKeyExtractor.java | 22 +- .../paimon/operation/FileStoreCommitTest.java | 42 +++ .../sink/FixedBucketRowKeyExtractorTest.java | 47 +++- .../sink/PartitionBucketMappingTest.java | 66 +++++ .../CdcRecordKeyAndBucketExtractorTest.java | 21 +- .../paimon/flink/action/CompactAction.java | 15 +- .../paimon/flink/action/RescaleAction.java | 6 +- .../paimon/flink/sink/FlinkSinkBuilder.java | 5 +- .../flink/sink/RowDataChannelComputer.java | 11 +- .../sink/RowDataKeyAndBucketExtractor.java | 57 ---- .../paimon/flink/ReadWriteTableITCase.java | 34 ++- .../flink/action/RescaleActionITCase.java | 244 ++++++++++++++++++ .../sink/RowDataChannelComputerTest.java | 8 +- 27 files changed, 813 insertions(+), 148 deletions(-) create mode 100644 paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java create mode 100644 paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java create mode 100644 paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java delete mode 100644 paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataKeyAndBucketExtractor.java create mode 100644 paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/action/RescaleActionITCase.java diff --git a/docs/docs/flink/procedures.md b/docs/docs/flink/procedures.md index dc4f1fedaf46..cc462d4bda74 100644 --- a/docs/docs/flink/procedures.md +++ b/docs/docs/flink/procedures.md @@ -867,7 +867,7 @@ All available procedures are listed below. CALL [catalog.]sys.rescale(`table` => 'identifier', `bucket_num` => bucket_num, `partition` => 'partition', `scan_parallelism` => scan_parallelism, `sink_parallelism` => sink_parallelism) - Rescale one partition of a table. Arguments: + Rescale one partition of a table. For partitioned tables, different partitions can have different bucket counts after rescaling. Arguments:
  • table: The target table identifier. Cannot be empty.
  • bucket_num: Resulting bucket number after rescale. The default value of argument bucket_num is the current bucket number of the table. Cannot be empty for postpone bucket tables.
  • partition: What partition to rescale. For partitioned table this argument cannot be empty.
    • diff --git a/docs/docs/maintenance/rescale-bucket.md b/docs/docs/maintenance/rescale-bucket.md index 1a304525345e..74889941f7ff 100644 --- a/docs/docs/maintenance/rescale-bucket.md +++ b/docs/docs/maintenance/rescale-bucket.md @@ -45,14 +45,17 @@ Please note that - `ALTER TABLE` only modifies the table's metadata and will **NOT** reorganize or reformat existing data. Reorganize existing data must be achieved by `INSERT OVERWRITE`. - Rescale bucket number does not influence the read and running write jobs. -- Once the bucket number is changed, any newly scheduled `INSERT INTO` jobs which write to without-reorganized - existing table/partition will throw a `TableException` with message like +- **Partitioned tables** support per-partition bucket counts. Each partition retains its own bucket + count from its data files, and the new bucket count only applies to newly created partitions or partitions that + have been reorganized with `INSERT OVERWRITE`. +- **Unpartitioned tables** require a full rescale before writing. If you change the bucket number and attempt + to write without reorganizing the data first, a `RuntimeException` will be thrown: ```text - Try to write table/partition ... with a new bucket num ..., + Try to write table with a new bucket num ..., but the previous bucket num is ... Please switch to batch mode, and perform INSERT OVERWRITE to rescale current data layout first. ``` -- For partitioned table, it is possible to have different bucket number for different partitions. *E.g.* +- For partitioned tables, it is possible to have different bucket numbers for different partitions. *E.g.* ```sql ALTER TABLE my_table SET ('bucket' = '4'); INSERT OVERWRITE my_table PARTITION (dt = '2022-01-01') @@ -62,6 +65,8 @@ Please note that INSERT OVERWRITE my_table PARTITION (dt = '2022-01-02') SELECT * FROM ...; ``` + After these operations, partition `dt=2022-01-01` uses 4 buckets, `dt=2022-01-02` uses 8 buckets, and any + new partitions will use the latest table-level default (8 buckets in this case). - During overwrite period, make sure there are no other jobs writing the same table/partition. ## Use Case @@ -121,8 +126,12 @@ and the job's latency keeps increasing. To improve the data freshness, users can -- scaling out ALTER TABLE verified_orders SET ('bucket' = '32'); ``` -- Switch to the batch mode and overwrite the current partition(s) to which the streaming job is writing +- Use the `rescale` procedure or switch to batch mode and overwrite the partition(s) that need rescaling ```sql + -- Option 1: Use the rescale procedure (recommended) + CALL sys.rescale(`table` => 'default.verified_orders', `bucket_num` => 32, `partition` => 'dt=2022-06-22'); + + -- Option 2: Manual batch overwrite SET 'execution.runtime-mode' = 'batch'; -- suppose today is 2022-06-22 -- case 1: there is no late event which updates the historical partitions, thus overwrite today's partition is enough @@ -142,8 +151,11 @@ and the job's latency keeps increasing. To improve the data freshness, users can FROM verified_orders WHERE dt IN ('2022-06-20', '2022-06-21', '2022-06-22'); ``` -- After overwrite job has finished, switch back to streaming mode. And now, the parallelism can be increased alongside with bucket number to restore the streaming job from the savepoint -( see [Start a SQL Job from a savepoint](https://nightlies.apache.org/flink/flink-docs-stable/docs/dev/table/sqlclient/#start-a-sql-job-from-a-savepoint) ) +- After the overwrite job has finished, switch back to streaming mode. The parallelism can be increased alongside + the bucket number to restore the streaming job from the savepoint + ( see [Start a SQL Job from a savepoint](https://nightlies.apache.org/flink/flink-docs-stable/docs/dev/table/sqlclient/#start-a-sql-job-from-a-savepoint) ). + Note that for partitioned tables, each partition retains its own bucket count, so only the rescaled partitions + are affected. ```sql SET 'execution.runtime-mode' = 'streaming'; SET 'execution.savepoint.path' = ; diff --git a/docs/docs/primary-key-table/data-distribution.md b/docs/docs/primary-key-table/data-distribution.md index 95e8c01aa6c9..643a287c0c27 100644 --- a/docs/docs/primary-key-table/data-distribution.md +++ b/docs/docs/primary-key-table/data-distribution.md @@ -24,7 +24,7 @@ under the License. # Data Distribution -A bucket is the smallest storage unit for reads and writes, each bucket directory contains an [LSM tree](./#lsm-trees). +A bucket is the smallest storage unit for reads and writes, each bucket directory contains an [LSM tree](./overview#lsm-trees). ## Fixed Bucket @@ -34,6 +34,9 @@ the bucket of record. Rescaling buckets can only be done through offline processes, see [Rescale Bucket](../maintenance/rescale-bucket). A too large number of buckets leads to too many small files, and a too small number of buckets leads to poor write performance. +For partitioned tables, each partition can have its own bucket count. After a rescale operation, existing +partitions retain their original bucket count while newly created partitions use the updated table-level default. + ## Dynamic Bucket Default mode for primary key table, or configure `'bucket' = '-1'`. diff --git a/paimon-core/src/main/java/org/apache/paimon/operation/AbstractFileStoreWrite.java b/paimon-core/src/main/java/org/apache/paimon/operation/AbstractFileStoreWrite.java index a49fff06d0ce..66a6ed76ce72 100644 --- a/paimon-core/src/main/java/org/apache/paimon/operation/AbstractFileStoreWrite.java +++ b/paimon-core/src/main/java/org/apache/paimon/operation/AbstractFileStoreWrite.java @@ -508,20 +508,28 @@ private RestoreFiles scanExistingFileMetas(BinaryRow partition, int bucket) { totalBuckets = restoredTotalBuckets; } if (!ignoreNumBucketCheck && totalBuckets != numBuckets) { - String partInfo = - partitionType.getFieldCount() > 0 - ? "partition " - + getPartitionComputer( - partitionType, - PARTITION_DEFAULT_NAME.defaultValue(), - legacyPartitionName) - .generatePartValues(partition) - : "table"; - throw new RuntimeException( - String.format( - "Try to write %s with a new bucket num %d, but the previous bucket num is %d. " - + "Please switch to batch mode, and perform INSERT OVERWRITE to rescale current data layout first.", - partInfo, numBuckets, totalBuckets)); + if (partitionType.getFieldCount() > 0) { + // For partitioned tables, allow per-partition bucket counts. + // The partition's existing bucket count takes precedence over the + // table-level default. This supports rescale operations where different + // partitions may have different bucket counts. + LOG.info( + "Partition {} uses {} buckets (table default: {}). " + + "Accepting per-partition bucket count.", + getPartitionComputer( + partitionType, + PARTITION_DEFAULT_NAME.defaultValue(), + legacyPartitionName) + .generatePartValues(partition), + totalBuckets, + numBuckets); + } else { + throw new RuntimeException( + String.format( + "Try to write table with a new bucket num %d, but the previous bucket num is %d. " + + "Please switch to batch mode, and perform INSERT OVERWRITE to rescale current data layout first.", + numBuckets, totalBuckets)); + } } return restored; } diff --git a/paimon-core/src/main/java/org/apache/paimon/privilege/PrivilegedFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/privilege/PrivilegedFileStoreTable.java index 35dfd308df0a..831bfb44eedf 100644 --- a/paimon-core/src/main/java/org/apache/paimon/privilege/PrivilegedFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/privilege/PrivilegedFileStoreTable.java @@ -27,6 +27,7 @@ import org.apache.paimon.table.ExpireSnapshots; import org.apache.paimon.table.FileStoreTable; import org.apache.paimon.table.query.LocalTableQuery; +import org.apache.paimon.table.sink.RowKeyExtractor; import org.apache.paimon.table.sink.TableCommitImpl; import org.apache.paimon.table.sink.TableWriteImpl; import org.apache.paimon.table.sink.WriteSelector; @@ -258,6 +259,13 @@ public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) return wrapped.newWrite(commitUser, writeId); } + @Override + public TableWriteImpl newWrite( + String commitUser, @Nullable Integer writeId, RowKeyExtractor rowKeyExtractor) { + privilegeChecker.assertCanInsert(identifier); + return wrapped.newWrite(commitUser, writeId, rowKeyExtractor); + } + @Override public TableCommitImpl newCommit(String commitUser) { privilegeChecker.assertCanInsert(identifier); diff --git a/paimon-core/src/main/java/org/apache/paimon/table/AbstractFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/AbstractFileStoreTable.java index df4224915927..03f73e8a5a34 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/AbstractFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/AbstractFileStoreTable.java @@ -40,6 +40,7 @@ import org.apache.paimon.table.sink.DynamicBucketRowKeyExtractor; import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; import org.apache.paimon.table.sink.FixedBucketWriteSelector; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.table.sink.PostponeBucketRowKeyExtractor; import org.apache.paimon.table.sink.RowKeyExtractor; import org.apache.paimon.table.sink.RowKindGenerator; @@ -230,7 +231,9 @@ public Optional statistics() { public Optional newWriteSelector() { switch (bucketMode()) { case HASH_FIXED: - return Optional.of(new FixedBucketWriteSelector(schema())); + return Optional.of( + new FixedBucketWriteSelector( + schema(), PartitionBucketMapping.loadFromTable(this))); case BUCKET_UNAWARE: case POSTPONE_MODE: return Optional.empty(); @@ -258,7 +261,8 @@ protected CatalogEnvironment newCatalogEnvironment(String branch) { public RowKeyExtractor createRowKeyExtractor() { switch (bucketMode()) { case HASH_FIXED: - return new FixedBucketRowKeyExtractor(schema()); + return new FixedBucketRowKeyExtractor( + schema(), PartitionBucketMapping.loadFromTable(this)); case HASH_DYNAMIC: case KEY_DYNAMIC: return new DynamicBucketRowKeyExtractor(schema()); diff --git a/paimon-core/src/main/java/org/apache/paimon/table/AppendOnlyFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/AppendOnlyFileStoreTable.java index d65c84fd5e65..f2f5280c3211 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/AppendOnlyFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/AppendOnlyFileStoreTable.java @@ -29,6 +29,7 @@ import org.apache.paimon.predicate.Predicate; import org.apache.paimon.schema.TableSchema; import org.apache.paimon.table.query.LocalTableQuery; +import org.apache.paimon.table.sink.RowKeyExtractor; import org.apache.paimon.table.sink.TableWriteImpl; import org.apache.paimon.table.source.AppendOnlySplitGenerator; import org.apache.paimon.table.source.AppendTableRead; @@ -140,11 +141,17 @@ public TableWriteImpl newWrite(String commitUser) { @Override public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) { + return newWrite(commitUser, writeId, createRowKeyExtractor()); + } + + @Override + public TableWriteImpl newWrite( + String commitUser, @Nullable Integer writeId, RowKeyExtractor rowKeyExtractor) { BaseAppendFileStoreWrite writer = store().newWrite(commitUser, writeId); return new TableWriteImpl<>( rowType(), writer, - createRowKeyExtractor(), + rowKeyExtractor, (record, rowKind) -> { Preconditions.checkState( rowKind.isAdd(), diff --git a/paimon-core/src/main/java/org/apache/paimon/table/DelegatedFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/DelegatedFileStoreTable.java index 7ba4bc20a9d7..76f222719db3 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/DelegatedFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/DelegatedFileStoreTable.java @@ -340,6 +340,12 @@ public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) return wrapped.newWrite(commitUser, writeId); } + @Override + public TableWriteImpl newWrite( + String commitUser, @Nullable Integer writeId, RowKeyExtractor rowKeyExtractor) { + return wrapped.newWrite(commitUser, writeId, rowKeyExtractor); + } + @Override public TableCommitImpl newCommit(String commitUser) { return wrapped.newCommit(commitUser); diff --git a/paimon-core/src/main/java/org/apache/paimon/table/FileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/FileStoreTable.java index b07465a25828..0b284bf5dd4d 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/FileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/FileStoreTable.java @@ -121,6 +121,13 @@ default Optional comment() { TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId); + /** + * Create a new write with a custom {@link RowKeyExtractor}. This is useful for scenarios like + * rescaling where the bucket assignment logic needs to be overridden. + */ + TableWriteImpl newWrite( + String commitUser, @Nullable Integer writeId, RowKeyExtractor rowKeyExtractor); + @Override TableCommitImpl newCommit(String commitUser); diff --git a/paimon-core/src/main/java/org/apache/paimon/table/PrimaryKeyFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/PrimaryKeyFileStoreTable.java index a2fee49bfb88..b68e5e48c8f2 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/PrimaryKeyFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/PrimaryKeyFileStoreTable.java @@ -32,6 +32,7 @@ import org.apache.paimon.schema.KeyValueFieldsExtractor; import org.apache.paimon.schema.TableSchema; import org.apache.paimon.table.query.LocalTableQuery; +import org.apache.paimon.table.sink.RowKeyExtractor; import org.apache.paimon.table.sink.TableWriteImpl; import org.apache.paimon.table.source.InnerTableRead; import org.apache.paimon.table.source.KeyValueTableRead; @@ -157,11 +158,18 @@ public TableWriteImpl newWrite(String commitUser) { @Override public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) { + return newWrite(commitUser, writeId, createRowKeyExtractor()); + } + + @Override + public TableWriteImpl newWrite( + String commitUser, @Nullable Integer writeId, RowKeyExtractor rowKeyExtractor) { + KeyValue kv = new KeyValue(); return new TableWriteImpl<>( rowType(), store().newWrite(commitUser, writeId), - createRowKeyExtractor(), + rowKeyExtractor, (record, rowKind) -> kv.replace( record.primaryKey(), diff --git a/paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java new file mode 100644 index 000000000000..4d2244f8640b --- /dev/null +++ b/paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java @@ -0,0 +1,92 @@ +/* + * 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.paimon.table; + +import org.apache.paimon.schema.TableSchema; +import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; +import org.apache.paimon.table.sink.FixedBucketWriteSelector; +import org.apache.paimon.table.sink.PartitionBucketMapping; +import org.apache.paimon.table.sink.RowKeyExtractor; +import org.apache.paimon.table.sink.TableWriteImpl; +import org.apache.paimon.table.sink.WriteSelector; + +import javax.annotation.Nullable; + +import java.util.Map; +import java.util.Optional; + +/** + * A {@link FileStoreTable} wrapper for rescale operations that overrides bucket-related behavior to + * use the new target bucket count instead of loading per-partition bucket mappings from the + * manifest. + */ +public class RescaleFileStoreTable extends DelegatedFileStoreTable { + + public RescaleFileStoreTable(FileStoreTable wrapped) { + super(wrapped); + } + + @Override + public Optional newWriteSelector() { + return Optional.of( + new FixedBucketWriteSelector( + schema(), new PartitionBucketMapping(schema().numBuckets()))); + } + + @Override + public RowKeyExtractor createRowKeyExtractor() { + return new FixedBucketRowKeyExtractor( + schema(), new PartitionBucketMapping(schema().numBuckets())); + } + + @Override + public TableWriteImpl newWrite(String commitUser) { + return newWrite(commitUser, null); + } + + @Override + public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) { + return wrapped().newWrite(commitUser, writeId, createRowKeyExtractor()); + } + + @Override + public FileStoreTable copy(Map dynamicOptions) { + return new RescaleFileStoreTable(wrapped().copy(dynamicOptions)); + } + + @Override + public FileStoreTable copy(TableSchema newTableSchema) { + return new RescaleFileStoreTable(wrapped().copy(newTableSchema)); + } + + @Override + public FileStoreTable copyWithoutTimeTravel(Map dynamicOptions) { + return new RescaleFileStoreTable(wrapped().copyWithoutTimeTravel(dynamicOptions)); + } + + @Override + public FileStoreTable copyWithLatestSchema() { + return new RescaleFileStoreTable(wrapped().copyWithLatestSchema()); + } + + @Override + public FileStoreTable switchToBranch(String branchName) { + return new RescaleFileStoreTable(wrapped().switchToBranch(branchName)); + } +} diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractor.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractor.java index 146a45b43713..aaf677eb48b4 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractor.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractor.java @@ -29,24 +29,27 @@ /** {@link KeyAndBucketExtractor} for {@link InternalRow}. */ public class FixedBucketRowKeyExtractor extends RowKeyExtractor { - private final int numBuckets; + private transient Projection bucketKeyProjection; + private final boolean sameBucketKeyAndTrimmedPrimaryKey; - private final Projection bucketKeyProjection; + private final PartitionBucketMapping partitionBucketMapping; private BinaryRow reuseBucketKey; private Integer reuseBucket; private final BucketFunction bucketFunction; - public FixedBucketRowKeyExtractor(TableSchema schema) { + public FixedBucketRowKeyExtractor( + TableSchema schema, PartitionBucketMapping partitionBucketMapping) { super(schema); - numBuckets = new CoreOptions(schema.options()).bucket(); - bucketFunction = - BucketFunction.create( - new CoreOptions(schema.options()), schema.logicalBucketKeyType()); - sameBucketKeyAndTrimmedPrimaryKey = schema.bucketKeys().equals(schema.trimmedPrimaryKeys()); - bucketKeyProjection = - CodeGenUtils.newProjection( - schema.logicalRowType(), schema.projection(schema.bucketKeys())); + this.bucketFunction = createBucketFunction(schema); + this.sameBucketKeyAndTrimmedPrimaryKey = + schema.bucketKeys().equals(schema.trimmedPrimaryKeys()); + this.partitionBucketMapping = partitionBucketMapping; + } + + private static BucketFunction createBucketFunction(TableSchema schema) { + return BucketFunction.create( + new CoreOptions(schema.options()), schema.logicalBucketKeyType()); } @Override @@ -62,7 +65,7 @@ private BinaryRow bucketKey() { } if (reuseBucketKey == null) { - reuseBucketKey = bucketKeyProjection.apply(record); + reuseBucketKey = bucketKeyProjection().apply(record); } return reuseBucketKey; } @@ -70,6 +73,7 @@ private BinaryRow bucketKey() { @Override public int bucket() { if (reuseBucket == null) { + int numBuckets = partitionBucketMapping.resolveNumBuckets(partition()); reuseBucket = bucket(numBuckets); } return reuseBucket; @@ -78,4 +82,13 @@ public int bucket() { public int bucket(int numBuckets) { return bucketFunction.bucket(bucketKey(), numBuckets); } + + private Projection bucketKeyProjection() { + if (bucketKeyProjection == null) { + bucketKeyProjection = + CodeGenUtils.newProjection( + schema.logicalRowType(), schema.projection(schema.bucketKeys())); + } + return bucketKeyProjection; + } } diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketWriteSelector.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketWriteSelector.java index e08841dd8cd3..a53bd7efb40c 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketWriteSelector.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/FixedBucketWriteSelector.java @@ -28,17 +28,20 @@ public class FixedBucketWriteSelector implements WriteSelector { private static final long serialVersionUID = 1L; private final TableSchema schema; + private final PartitionBucketMapping partitionBucketMapping; private transient KeyAndBucketExtractor extractor; - public FixedBucketWriteSelector(TableSchema schema) { + public FixedBucketWriteSelector( + TableSchema schema, PartitionBucketMapping partitionBucketMapping) { this.schema = schema; + this.partitionBucketMapping = partitionBucketMapping; } @Override public int select(InternalRow row, int numWriters) { if (extractor == null) { - extractor = new FixedBucketRowKeyExtractor(schema); + extractor = new FixedBucketRowKeyExtractor(schema, partitionBucketMapping); } extractor.setRecord(row); return ChannelComputer.select(extractor.partition(), extractor.bucket(), numWriters); diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java new file mode 100644 index 000000000000..7e2a9d87db9b --- /dev/null +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java @@ -0,0 +1,130 @@ +/* + * 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.paimon.table.sink; + +import org.apache.paimon.data.BinaryRow; +import org.apache.paimon.manifest.ManifestEntry; +import org.apache.paimon.table.FileStoreTable; + +import java.io.Serializable; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +/** + * A mapping that resolves the number of buckets for each partition in a table. + * + *

    Different partitions may have different bucket counts (e.g., after a rescale operation). This + * class maintains a per-partition bucket count mapping and falls back to a default bucket count for + * partitions that are not explicitly mapped. + * + *

    This is used by components such as {@link FixedBucketRowKeyExtractor} and {@link + * FixedBucketWriteSelector} to correctly determine the bucket assignment for rows in tables where + * partitions may have been rescaled independently. + * + * @see #loadFromTable(FileStoreTable) + * @see #resolveNumBuckets(BinaryRow) + */ +public class PartitionBucketMapping implements Serializable { + + private static final long serialVersionUID = 1L; + + /** The default number of buckets, used when a partition has no explicit mapping. */ + private final int defaultBucketCount; + + /** A map from partition to its specific bucket count. May be empty but never {@code null}. */ + private final Map partitionBucketMap; + + /** + * Creates a mapping with only a default bucket count and no per-partition overrides. + * + * @param defaultBucketCount the default number of buckets for all partitions + */ + public PartitionBucketMapping(int defaultBucketCount) { + this(defaultBucketCount, Collections.emptyMap()); + } + + /** + * Creates a mapping with a default bucket count and an explicit per-partition bucket map. + * + * @param defaultBucketCount the default number of buckets, used as a fallback + * @param partitionBucketMap a map from partition (as {@link BinaryRow}) to its bucket count + */ + public PartitionBucketMapping( + int defaultBucketCount, Map partitionBucketMap) { + this.defaultBucketCount = defaultBucketCount; + this.partitionBucketMap = partitionBucketMap; + } + + /** + * Loads a {@link PartitionBucketMapping} by scanning the manifest entries of the given table. + * + *

    For non-partitioned tables, this returns a mapping with only the schema-defined default + * bucket count and an empty partition map. + * + *

    For partitioned tables, the method scans all manifest entries and records the {@code + * totalBuckets} value for each partition. If the scan fails for any reason, a fallback mapping + * with only the default bucket count is returned. + * + * @param table the {@link FileStoreTable} to load the mapping from + * @return a {@link PartitionBucketMapping} reflecting the current bucket layout of the table + */ + public static PartitionBucketMapping loadFromTable(FileStoreTable table) { + int defaultBuckets = table.schema().numBuckets(); + if (table.partitionKeys().isEmpty()) { + return new PartitionBucketMapping(defaultBuckets, Collections.emptyMap()); + } + + try { + List entries = table.store().newScan().plan().files(); + Map partitionBucketMap = new HashMap<>(); + for (ManifestEntry entry : entries) { + int totalBuckets = entry.totalBuckets(); + if (totalBuckets > 0) { + BinaryRow partition = entry.partition(); + partitionBucketMap.putIfAbsent(partition.copy(), totalBuckets); + } + } + + return new PartitionBucketMapping(defaultBuckets, partitionBucketMap); + } catch (Exception e) { + return new PartitionBucketMapping(defaultBuckets, Collections.emptyMap()); + } + } + + /** + * Resolves the number of buckets for the given partition. + * + *

    If the partition has an explicit entry in the partition-to-bucket map, that value is + * returned. Otherwise, the default bucket count is returned. + * + * @param partition the partition key as a {@link BinaryRow} + * @return the number of buckets for the given partition + */ + public int resolveNumBuckets(BinaryRow partition) { + if (partitionBucketMap != null) { + Integer partitionBucketCount = partitionBucketMap.get(partition); + if (partitionBucketCount != null) { + return partitionBucketCount; + } + } + return defaultBucketCount; + } +} diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/RowKeyExtractor.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/RowKeyExtractor.java index 455aaa4aa5e9..697734ca10b1 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/sink/RowKeyExtractor.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/RowKeyExtractor.java @@ -22,18 +22,23 @@ import org.apache.paimon.data.InternalRow; import org.apache.paimon.schema.TableSchema; +import java.io.Serializable; + /** {@link KeyAndBucketExtractor} for {@link InternalRow}. */ -public abstract class RowKeyExtractor implements KeyAndBucketExtractor { +public abstract class RowKeyExtractor implements KeyAndBucketExtractor, Serializable { + + private static final long serialVersionUID = 1L; - private final RowPartitionKeyExtractor partitionKeyExtractor; + private transient RowPartitionKeyExtractor partitionKeyExtractor; + protected final TableSchema schema; protected InternalRow record; private BinaryRow partition; private BinaryRow trimmedPrimaryKey; public RowKeyExtractor(TableSchema schema) { - this.partitionKeyExtractor = new RowPartitionKeyExtractor(schema); + this.schema = schema; } @Override @@ -46,7 +51,7 @@ public void setRecord(InternalRow record) { @Override public BinaryRow partition() { if (partition == null) { - partition = partitionKeyExtractor.partition(record); + partition = partitionKeyExtractor().partition(record); } return partition; } @@ -54,8 +59,15 @@ public BinaryRow partition() { @Override public BinaryRow trimmedPrimaryKey() { if (trimmedPrimaryKey == null) { - trimmedPrimaryKey = partitionKeyExtractor.trimmedPrimaryKey(record); + trimmedPrimaryKey = partitionKeyExtractor().trimmedPrimaryKey(record); } return trimmedPrimaryKey; } + + private RowPartitionKeyExtractor partitionKeyExtractor() { + if (partitionKeyExtractor == null) { + partitionKeyExtractor = new RowPartitionKeyExtractor(schema); + } + return partitionKeyExtractor; + } } diff --git a/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java b/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java index 71eb081de89f..2efb5ab4974e 100644 --- a/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java @@ -1080,6 +1080,48 @@ public void testCommitRetryAfterFalseSuccessDoesNotCleanManifest() throws Except assertThat(store.readKvsFromSnapshot(latestSnapshot.id())).hasSize(1); } + @Test + public void testBucketCountConsistencyValidation() throws Exception { + TestFileStore store = createStore(false); + + // Commit initial data + List data = generateDataList(10); + store.commitData(data, gen::getPartition, kv -> 0); + + // Re-commit the same data but with a different totalBuckets value. + // This simulates a stale writer that loaded an old bucket mapping. + assertThatThrownBy( + () -> + store.commitDataImpl( + data, + gen::getPartition, + kv -> 0, + false, + null, + null, + Collections.emptyList(), + (commit, committable) -> { + ManifestCommittable tampered = + new ManifestCommittable( + committable.identifier(), + committable.watermark()); + for (CommitMessage msg : + committable.fileCommittables()) { + CommitMessageImpl impl = (CommitMessageImpl) msg; + tampered.addFileCommittable( + new CommitMessageImpl( + impl.partition(), + impl.bucket(), + 99, + impl.newFilesIncrement(), + impl.compactIncrement())); + } + commit.commit(tampered, true); + })) + .isInstanceOf(RuntimeException.class) + .hasMessageContaining("without overwrite"); + } + private FileStoreCommitImpl newCommitWithSnapshotCommit( TestFileStore store, String commitUser, SnapshotCommit snapshotCommit) { String tableName = store.options().path().getName(); diff --git a/paimon-core/src/test/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractorTest.java b/paimon-core/src/test/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractorTest.java index 5c551dac7004..82a1a53d0a41 100644 --- a/paimon-core/src/test/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractorTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/table/sink/FixedBucketRowKeyExtractorTest.java @@ -101,6 +101,30 @@ public void testUnCompactDecimalAndTimestampNullValueBucketNumber() { } } + @Test + public void testPerPartitionBucketCount() { + int defaultBuckets = 100; + int partition1Buckets = 4; + + // Build a BinaryRow for partition value = 1 + BinaryRow partitionRow = BinaryRow.singleColumn(1); + + Map partitionMap = new HashMap<>(); + partitionMap.put(partitionRow, partition1Buckets); + PartitionBucketMapping mapping = new PartitionBucketMapping(defaultBuckets, partitionMap); + + // Schema: partition key "a", bucket key "b", primary key "a,b" + FixedBucketRowKeyExtractor extractor = extractor("a", "b", "a,b", defaultBuckets, mapping); + + // Same bucket key (b=456) in both partitions, different bucket counts produce + // different bucket assignments: hash(456) % 4 = 3, hash(456) % 100 = 47 + GenericRow rowInMappedPartition = GenericRow.of(1, 456, 7); + assertThat(bucket(extractor, rowInMappedPartition)).isEqualTo(3); + + GenericRow rowInDefaultPartition = GenericRow.of(99, 456, 7); + assertThat(bucket(extractor, rowInDefaultPartition)).isEqualTo(47); + } + private int bucket(FixedBucketRowKeyExtractor extractor, InternalRow row) { extractor.setRecord(row); return extractor.bucket(); @@ -125,8 +149,29 @@ private FixedBucketRowKeyExtractor extractor( return extractor(rowType, partK, bk, pk, numBucket); } + private FixedBucketRowKeyExtractor extractor( + String partK, String bk, String pk, int numBucket, PartitionBucketMapping mapping) { + RowType rowType = + new RowType( + Arrays.asList( + new DataField(0, "a", new IntType()), + new DataField(1, "b", new IntType()), + new DataField(2, "c", new IntType()))); + return extractor(rowType, partK, bk, pk, numBucket, mapping); + } + private FixedBucketRowKeyExtractor extractor( RowType rowType, String partK, String bk, String pk, int numBucket) { + return extractor(rowType, partK, bk, pk, numBucket, new PartitionBucketMapping(numBucket)); + } + + private FixedBucketRowKeyExtractor extractor( + RowType rowType, + String partK, + String bk, + String pk, + int numBucket, + PartitionBucketMapping mapping) { List fields = TableSchema.newFields(rowType); Map options = new HashMap<>(); options.put(BUCKET_KEY.key(), bk); @@ -142,6 +187,6 @@ private FixedBucketRowKeyExtractor extractor( "".equals(pk) ? Collections.emptyList() : Arrays.asList(pk.split(",")), options, ""); - return new FixedBucketRowKeyExtractor(schema); + return new FixedBucketRowKeyExtractor(schema, mapping); } } diff --git a/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java b/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java new file mode 100644 index 000000000000..f6958720c30f --- /dev/null +++ b/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java @@ -0,0 +1,66 @@ +/* + * 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.paimon.table.sink; + +import org.apache.paimon.data.BinaryRow; + +import org.junit.jupiter.api.Test; + +import java.util.HashMap; +import java.util.Map; + +import static org.assertj.core.api.Assertions.assertThat; + +/** Tests for {@link PartitionBucketMapping}. */ +public class PartitionBucketMappingTest { + + @Test + public void testDefaultBucketCount() { + PartitionBucketMapping mapping = new PartitionBucketMapping(16); + + // Any partition should resolve to the default + assertThat(mapping.resolveNumBuckets(BinaryRow.EMPTY_ROW)).isEqualTo(16); + assertThat(mapping.resolveNumBuckets(partition(1))).isEqualTo(16); + assertThat(mapping.resolveNumBuckets(partition(42))).isEqualTo(16); + } + + @Test + public void testExplicitPartitionMapping() { + BinaryRow partA = partition(1); + BinaryRow partB = partition(2); + BinaryRow partC = partition(3); + + Map partitionMap = new HashMap<>(); + partitionMap.put(partA, 32); + partitionMap.put(partB, 64); + + PartitionBucketMapping mapping = new PartitionBucketMapping(16, partitionMap); + + // Mapped partitions return their specific bucket counts + assertThat(mapping.resolveNumBuckets(partA)).isEqualTo(32); + assertThat(mapping.resolveNumBuckets(partB)).isEqualTo(64); + + // Unmapped partition falls back to the default + assertThat(mapping.resolveNumBuckets(partC)).isEqualTo(16); + } + + private static BinaryRow partition(int value) { + return BinaryRow.singleColumn(value); + } +} diff --git a/paimon-flink/paimon-flink-cdc/src/test/java/org/apache/paimon/flink/sink/cdc/CdcRecordKeyAndBucketExtractorTest.java b/paimon-flink/paimon-flink-cdc/src/test/java/org/apache/paimon/flink/sink/cdc/CdcRecordKeyAndBucketExtractorTest.java index 802a3ea9d4cf..642ef76ec219 100644 --- a/paimon-flink/paimon-flink-cdc/src/test/java/org/apache/paimon/flink/sink/cdc/CdcRecordKeyAndBucketExtractorTest.java +++ b/paimon-flink/paimon-flink-cdc/src/test/java/org/apache/paimon/flink/sink/cdc/CdcRecordKeyAndBucketExtractorTest.java @@ -18,13 +18,15 @@ package org.apache.paimon.flink.sink.cdc; -import org.apache.paimon.flink.sink.RowDataKeyAndBucketExtractor; +import org.apache.paimon.flink.FlinkRowWrapper; import org.apache.paimon.fs.Path; import org.apache.paimon.fs.local.LocalFileIO; import org.apache.paimon.schema.Schema; import org.apache.paimon.schema.SchemaManager; import org.apache.paimon.schema.SchemaUtils; import org.apache.paimon.schema.TableSchema; +import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.types.DataType; import org.apache.paimon.types.DataTypes; import org.apache.paimon.types.RowKind; @@ -65,7 +67,7 @@ public class CdcRecordKeyAndBucketExtractorTest { public void testExtract() throws Exception { ThreadLocalRandom random = ThreadLocalRandom.current(); TableSchema schema = createTableSchema(); - RowDataKeyAndBucketExtractor expected = new RowDataKeyAndBucketExtractor(schema); + FixedBucketRowKeyExtractor expected = extractor(schema); CdcRecordKeyAndBucketExtractor actual = new CdcRecordKeyAndBucketExtractor(schema); int numTests = random.nextInt(1000) + 1; @@ -85,7 +87,7 @@ public void testExtract() throws Exception { v1, StringData.fromString(k2), StringData.fromString(v2)); - expected.setRecord(rowData); + expected.setRecord(new FlinkRowWrapper(rowData)); Map data = new HashMap<>(); data.put("pt1", pt1); @@ -109,7 +111,7 @@ public void testExtract() throws Exception { public void testNullPartition() throws Exception { ThreadLocalRandom random = ThreadLocalRandom.current(); TableSchema schema = createTableSchema(); - RowDataKeyAndBucketExtractor expected = new RowDataKeyAndBucketExtractor(schema); + FixedBucketRowKeyExtractor expected = extractor(schema); CdcRecordKeyAndBucketExtractor actual = new CdcRecordKeyAndBucketExtractor(schema); long k1 = random.nextLong(); @@ -120,7 +122,7 @@ public void testNullPartition() throws Exception { GenericRowData rowData = GenericRowData.of( null, null, k1, v1, StringData.fromString(k2), StringData.fromString(v2)); - expected.setRecord(rowData); + expected.setRecord(new FlinkRowWrapper(rowData)); Map data = new HashMap<>(); data.put("pt1", null); @@ -143,7 +145,7 @@ public void testNullPartition() throws Exception { public void testEmptyPartition() throws Exception { ThreadLocalRandom random = ThreadLocalRandom.current(); TableSchema schema = createTableSchema(); - RowDataKeyAndBucketExtractor expected = new RowDataKeyAndBucketExtractor(schema); + FixedBucketRowKeyExtractor expected = extractor(schema); CdcRecordKeyAndBucketExtractor actual = new CdcRecordKeyAndBucketExtractor(schema); long k1 = random.nextLong(); @@ -159,7 +161,7 @@ public void testEmptyPartition() throws Exception { v1, StringData.fromString(k2), StringData.fromString(v2)); - expected.setRecord(rowData); + expected.setRecord(new FlinkRowWrapper(rowData)); Map data = new HashMap<>(); data.put("pt1", ""); @@ -188,4 +190,9 @@ private TableSchema createTableSchema() throws Exception { Collections.singletonMap("bucket", "1"), "")); } + + private FixedBucketRowKeyExtractor extractor(TableSchema schema) { + return new FixedBucketRowKeyExtractor( + schema, new PartitionBucketMapping(schema.numBuckets())); + } } diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java index 7c4676397523..63003cffafbc 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java @@ -45,6 +45,8 @@ import org.apache.paimon.predicate.PredicateProjectionConverter; import org.apache.paimon.table.BucketMode; import org.apache.paimon.table.FileStoreTable; +import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.types.RowType; import org.apache.paimon.utils.InternalRowPartitionComputer; import org.apache.paimon.utils.Pair; @@ -322,7 +324,7 @@ protected boolean buildForPostponeBucketCompaction( String commitUser = CoreOptions.createCommitUser(options); List> dataStreams = new ArrayList<>(); for (BinaryRow partition : partitions) { - int bucketNum = defaultBucketNum; + int partitionBucketNum = defaultBucketNum; Iterator it = table.newSnapshotReader() @@ -330,11 +332,11 @@ protected boolean buildForPostponeBucketCompaction( .onlyReadRealBuckets() .readFileIterator(); if (it.hasNext()) { - bucketNum = it.next().totalBuckets(); + partitionBucketNum = it.next().totalBuckets(); } bucketOptions = new HashMap<>(table.options()); - bucketOptions.put(CoreOptions.BUCKET.key(), String.valueOf(bucketNum)); + bucketOptions.put(CoreOptions.BUCKET.key(), String.valueOf(partitionBucketNum)); FileStoreTable realTable = table.copy(table.schema().copy(bucketOptions)); LinkedHashMap partitionSpec = @@ -346,13 +348,18 @@ protected boolean buildForPostponeBucketCompaction( partitionSpec, options.get(FlinkConnectorOptions.SCAN_PARALLELISM)); + PartitionBucketMapping partitionBucketMapping = + new PartitionBucketMapping(partitionBucketNum); + FixedBucketRowKeyExtractor extractor = + new FixedBucketRowKeyExtractor(realTable.schema(), partitionBucketMapping); + DataStream partitioned = FlinkStreamPartitioner.partition( FlinkSinkBuilder.mapToInternalRow( sourcePair.getLeft(), realTable.rowType(), table.catalogEnvironment().catalogContext()), - new RowDataChannelComputer(realTable.schema()), + new RowDataChannelComputer(extractor), null); FixedBucketSink sink = new FixedBucketSink(realTable, null); DataStream written = diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java index 3855a5115d43..976bfb302201 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java @@ -26,6 +26,7 @@ import org.apache.paimon.partition.PartitionPredicate; import org.apache.paimon.table.BucketMode; import org.apache.paimon.table.FileStoreTable; +import org.apache.paimon.table.RescaleFileStoreTable; import org.apache.paimon.utils.Preconditions; import org.apache.flink.api.common.RuntimeExecutionMode; @@ -120,8 +121,9 @@ public void build() throws Exception { } else { bucketOptions.put(CoreOptions.BUCKET.key(), String.valueOf(bucketNum)); } - FileStoreTable rescaledTable = - fileStoreTable.copy(fileStoreTable.schema().copy(bucketOptions)); + RescaleFileStoreTable rescaledTable = + new RescaleFileStoreTable( + fileStoreTable.copy(fileStoreTable.schema().copy(bucketOptions))); new FlinkSinkBuilder(rescaledTable) .overwrite(partition) .parallelism(sinkParallelism == null ? bucketNum : sinkParallelism) diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java index ae8013b7e709..430eeb180c9f 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java @@ -300,8 +300,9 @@ protected DataStreamSink buildForFixedBucket(DataStream input) { + " then the parallelism of writerOperator will be set to bucketNums."); parallelism = bucketNums; } - DataStream partitioned = - partition(input, new RowDataChannelComputer(table.schema()), parallelism); + RowDataChannelComputer channelComputer = + new RowDataChannelComputer(table.createRowKeyExtractor()); + DataStream partitioned = partition(input, channelComputer, parallelism); FixedBucketSink sink = new FixedBucketSink(table, overwritePartition); return sink.sinkFrom(partitioned); } diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataChannelComputer.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataChannelComputer.java index 1df93c82bcb1..f785bf564491 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataChannelComputer.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataChannelComputer.java @@ -20,9 +20,7 @@ import org.apache.paimon.data.BinaryRow; import org.apache.paimon.data.InternalRow; -import org.apache.paimon.schema.TableSchema; import org.apache.paimon.table.sink.ChannelComputer; -import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; import org.apache.paimon.table.sink.KeyAndBucketExtractor; /** {@link ChannelComputer} for {@link InternalRow}. */ @@ -30,19 +28,16 @@ public class RowDataChannelComputer implements ChannelComputer { private static final long serialVersionUID = 1L; - private final TableSchema schema; - private transient int numChannels; - private transient KeyAndBucketExtractor extractor; + private final KeyAndBucketExtractor extractor; - public RowDataChannelComputer(TableSchema schema) { - this.schema = schema; + public RowDataChannelComputer(KeyAndBucketExtractor extractor) { + this.extractor = extractor; } @Override public void setup(int numChannels) { this.numChannels = numChannels; - this.extractor = new FixedBucketRowKeyExtractor(schema); } @Override diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataKeyAndBucketExtractor.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataKeyAndBucketExtractor.java deleted file mode 100644 index 8082df005fe8..000000000000 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/RowDataKeyAndBucketExtractor.java +++ /dev/null @@ -1,57 +0,0 @@ -/* - * 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.paimon.flink.sink; - -import org.apache.paimon.data.BinaryRow; -import org.apache.paimon.flink.FlinkRowWrapper; -import org.apache.paimon.schema.TableSchema; -import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; -import org.apache.paimon.table.sink.KeyAndBucketExtractor; - -import org.apache.flink.table.data.RowData; - -/** {@link KeyAndBucketExtractor} for {@link RowData}. */ -public class RowDataKeyAndBucketExtractor implements KeyAndBucketExtractor { - - private final FixedBucketRowKeyExtractor wrapped; - - public RowDataKeyAndBucketExtractor(TableSchema schema) { - wrapped = new FixedBucketRowKeyExtractor(schema); - } - - @Override - public void setRecord(RowData record) { - wrapped.setRecord(new FlinkRowWrapper(record)); - } - - @Override - public BinaryRow partition() { - return wrapped.partition(); - } - - @Override - public int bucket() { - return wrapped.bucket(); - } - - @Override - public BinaryRow trimmedPrimaryKey() { - return wrapped.trimmedPrimaryKey(); - } -} diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/ReadWriteTableITCase.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/ReadWriteTableITCase.java index 940dd95af0b0..50ea1b5261e5 100644 --- a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/ReadWriteTableITCase.java +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/ReadWriteTableITCase.java @@ -1394,11 +1394,13 @@ public void testChangeBucketNumber() throws Exception { insertInto(table, "('US Dollar', 102, '2022-06-20')"); - // increase bucket num from 2 to 3 - assertChangeBucketWithoutRescale(table, 3); + // For partitioned tables, changing the bucket num and writing is allowed. + // The partition's existing bucket count takes precedence, supporting + // per-partition rescale operations. - // decrease bucket num from 3 to 1 - assertChangeBucketWithoutRescale(table, 1); + assertChangeBucketWithoutRescale(table, 3, 1); + + assertChangeBucketWithoutRescale(table, 1, 2); } @Test @@ -1938,7 +1940,7 @@ private void testSinkParallelism( // 3. assert parallelism from transformation DataStream mockSource = - bExeEnv.fromCollection(Collections.singletonList(GenericRowData.of())); + bExeEnv.fromData(Collections.singletonList(GenericRowData.of())); mockSource.getTransformation().setParallelism(mockSource.getParallelism(), false); DataStreamSink sink = sinkProvider.consumeDataStream(null, mockSource); @@ -1987,21 +1989,15 @@ private static Stream testSinkParallelismParameters() { return parameters.stream(); } - private void assertChangeBucketWithoutRescale(String table, int bucketNum) throws Exception { + private void assertChangeBucketWithoutRescale( + String table, int bucketNum, int expectedRowsBefore) throws Exception { bEnv.executeSql(String.format("ALTER TABLE `%s` SET ('bucket' = '%d')", table, bucketNum)); - // read is ok - assertThat( - BlockingIterator.of(bEnv.executeSql(buildSimpleQuery(table)).collect()) - .collect()) - .containsExactlyInAnyOrder(changelogRow("+I", "US Dollar", 102L, "2022-06-20")); - assertThatThrownBy(() -> insertInto(table, "('US Dollar', 102, '2022-06-20')")) - .rootCause() - .isInstanceOf(RuntimeException.class) - .hasMessage( - String.format( - "Try to write partition {dt=2022-06-20} with a new bucket num %d, but the previous bucket num is 2. " - + "Please switch to batch mode, and perform INSERT OVERWRITE to rescale current data layout first.", - bucketNum)); + // read existing data is ok after changing bucket num + List rows = + BlockingIterator.of(bEnv.executeSql(buildSimpleQuery(table)).collect()).collect(); + assertThat(rows).hasSize(expectedRowsBefore); + // writing with a different bucket num is allowed for partitioned tables + insertInto(table, "('US Dollar', 102, '2022-06-20')"); } private void validateSchemaOptionResult() { diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/action/RescaleActionITCase.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/action/RescaleActionITCase.java new file mode 100644 index 000000000000..4c698d23ff03 --- /dev/null +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/action/RescaleActionITCase.java @@ -0,0 +1,244 @@ +/* + * 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.paimon.flink.action; + +import org.apache.paimon.CoreOptions; +import org.apache.paimon.data.BinaryRow; +import org.apache.paimon.data.BinaryString; +import org.apache.paimon.data.InternalRow; +import org.apache.paimon.manifest.FileKind; +import org.apache.paimon.manifest.ManifestEntry; +import org.apache.paimon.table.FileStoreTable; +import org.apache.paimon.table.sink.StreamWriteBuilder; +import org.apache.paimon.types.DataType; +import org.apache.paimon.types.DataTypes; +import org.apache.paimon.types.RowType; + +import org.junit.jupiter.api.Test; +import org.junit.jupiter.api.Timeout; + +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +import static org.assertj.core.api.Assertions.assertThat; + +/** IT cases for {@link RescaleAction}. */ +public class RescaleActionITCase extends ActionITCaseBase { + + private static final RowType ROW_TYPE = + RowType.of( + new DataType[] {DataTypes.INT(), DataTypes.INT(), DataTypes.STRING()}, + new String[] {"k", "v", "pt"}); + + private FileStoreTable prepareTable(int initialBuckets) throws Exception { + Map options = new HashMap<>(); + options.put(CoreOptions.BUCKET.key(), String.valueOf(initialBuckets)); + FileStoreTable table = + createFileStoreTable( + ROW_TYPE, + Collections.singletonList("pt"), + Arrays.asList("pt", "k"), + Collections.emptyList(), + options); + + StreamWriteBuilder streamWriteBuilder = + table.newStreamWriteBuilder().withCommitUser(commitUser); + write = streamWriteBuilder.newWrite(); + commit = streamWriteBuilder.newCommit(); + return table; + } + + @Test + @Timeout(120) + public void testRescaleAllPartitions() throws Exception { + int initialBuckets = 2; + int newBuckets = 4; + prepareTable(initialBuckets); + + // Write data to two partitions + writeData( + rowData(1, 100, BinaryString.fromString("p1")), + rowData(2, 200, BinaryString.fromString("p1")), + rowData(3, 300, BinaryString.fromString("p2")), + rowData(4, 400, BinaryString.fromString("p2"))); + + // Rescale all partitions (no --partition argument) + createAction( + RescaleAction.class, + "rescale", + "--warehouse", + warehouse, + "--database", + database, + "--table", + tableName, + "--bucket_num", + String.valueOf(newBuckets)) + .run(); + + // Verify both partitions now have the new bucket count + FileStoreTable table = getFileStoreTable(tableName); + Map partitionBuckets = getPartitionBucketCounts(table); + for (Map.Entry entry : partitionBuckets.entrySet()) { + assertThat(entry.getValue()).isEqualTo(newBuckets); + } + + // Verify data is preserved + List rows = getData(tableName); + assertThat(rows).hasSize(4); + } + + @Test + @Timeout(120) + public void testRescaleSinglePartition() throws Exception { + int initialBuckets = 2; + int rescaledBuckets = 4; + prepareTable(initialBuckets); + + // Write data to two partitions + writeData( + rowData(1, 100, BinaryString.fromString("p1")), + rowData(2, 200, BinaryString.fromString("p1")), + rowData(3, 300, BinaryString.fromString("p2")), + rowData(4, 400, BinaryString.fromString("p2"))); + + // Rescale only partition p1 to 4 buckets, leaving p2 at 2 buckets + createAction( + RescaleAction.class, + "rescale", + "--warehouse", + warehouse, + "--database", + database, + "--table", + tableName, + "--bucket_num", + String.valueOf(rescaledBuckets), + "--partition", + "pt=p1") + .run(); + + // Verify the table now has different bucket counts per partition + FileStoreTable table = getFileStoreTable(tableName); + Map partitionBuckets = getPartitionBucketCounts(table); + + assertThat(partitionBuckets).hasSize(2); + + // Find which partition is p1 and which is p2 by checking their bucket counts + int p1Buckets = -1; + int p2Buckets = -1; + for (Map.Entry entry : partitionBuckets.entrySet()) { + int buckets = entry.getValue(); + // One partition should have rescaledBuckets, the other initialBuckets + if (buckets == rescaledBuckets) { + p1Buckets = buckets; + } else if (buckets == initialBuckets) { + p2Buckets = buckets; + } + } + assertThat(p1Buckets) + .as("Rescaled partition p1 should have %d buckets", rescaledBuckets) + .isEqualTo(rescaledBuckets); + assertThat(p2Buckets) + .as("Non-rescaled partition p2 should still have %d buckets", initialBuckets) + .isEqualTo(initialBuckets); + + // Verify data is preserved across both partitions + List rows = getData(tableName); + assertThat(rows).hasSize(4); + } + + @Test + @Timeout(120) + public void testRescaleAppendOnlyTable() throws Exception { + int initialBuckets = 2; + int newBuckets = 4; + + // Create an append-only table (no primary keys) + RowType appendRowType = + RowType.of( + new DataType[] {DataTypes.INT(), DataTypes.INT(), DataTypes.STRING()}, + new String[] {"a", "b", "pt"}); + Map options = new HashMap<>(); + options.put(CoreOptions.BUCKET.key(), String.valueOf(initialBuckets)); + FileStoreTable table = + createFileStoreTable( + appendRowType, + Collections.singletonList("pt"), + Collections.emptyList(), // no primary keys = append-only + Collections.singletonList("a"), // bucket key required for append-only + options); + + StreamWriteBuilder streamWriteBuilder = + table.newStreamWriteBuilder().withCommitUser(commitUser); + write = streamWriteBuilder.newWrite(); + commit = streamWriteBuilder.newCommit(); + + // Write data + writeData( + rowData(1, 100, BinaryString.fromString("p1")), + rowData(2, 200, BinaryString.fromString("p1")), + rowData(3, 300, BinaryString.fromString("p1")), + rowData(4, 400, BinaryString.fromString("p1"))); + + // Rescale the append-only table + createAction( + RescaleAction.class, + "rescale", + "--warehouse", + warehouse, + "--database", + database, + "--table", + tableName, + "--bucket_num", + String.valueOf(newBuckets)) + .run(); + + // Verify the new bucket count + FileStoreTable rescaledTable = getFileStoreTable(tableName); + Map partitionBuckets = getPartitionBucketCounts(rescaledTable); + for (Map.Entry entry : partitionBuckets.entrySet()) { + assertThat(entry.getValue()).isEqualTo(newBuckets); + } + + // Verify data is preserved + List rows = getData(tableName); + assertThat(rows).hasSize(4); + } + + /** + * Reads all manifest entries and builds a map from partition to totalBuckets. This reflects the + * actual bucket layout on disk. + */ + private Map getPartitionBucketCounts(FileStoreTable table) { + List entries = table.store().newScan().plan().files(FileKind.ADD); + Map result = new HashMap<>(); + for (ManifestEntry entry : entries) { + int totalBuckets = entry.totalBuckets(); + if (totalBuckets > 0) { + result.putIfAbsent(entry.partition().copy(), totalBuckets); + } + } + return result; + } +} diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/RowDataChannelComputerTest.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/RowDataChannelComputerTest.java index 687407141bf1..a4cf677481b8 100644 --- a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/RowDataChannelComputerTest.java +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/RowDataChannelComputerTest.java @@ -27,6 +27,7 @@ import org.apache.paimon.schema.SchemaManager; import org.apache.paimon.schema.TableSchema; import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.types.DataType; import org.apache.paimon.types.DataTypes; import org.apache.paimon.types.RowType; @@ -108,10 +109,13 @@ public void testSchemaNoPartition() throws Exception { private void testImpl(TableSchema schema, List input) { ThreadLocalRandom random = ThreadLocalRandom.current(); - FixedBucketRowKeyExtractor extractor = new FixedBucketRowKeyExtractor(schema); + PartitionBucketMapping partitionBucketMapping = + new PartitionBucketMapping(schema.numBuckets()); + FixedBucketRowKeyExtractor extractor = + new FixedBucketRowKeyExtractor(schema, partitionBucketMapping); int numChannels = random.nextInt(10) + 1; - RowDataChannelComputer channelComputer = new RowDataChannelComputer(schema); + RowDataChannelComputer channelComputer = new RowDataChannelComputer(extractor); channelComputer.setup(numChannels); // assert that channel(record) and channel(partition, bucket) gives the same result From f9b0a1fba5139d164c8d5e886aca0f7c318395ef Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Mon, 9 Mar 2026 12:39:20 +1100 Subject: [PATCH 2/7] Optimize PartitionBucketMapping.loadFromTable --- .../table/sink/PartitionBucketMapping.java | 22 +++++++++++-------- 1 file changed, 13 insertions(+), 9 deletions(-) diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java index 7e2a9d87db9b..bf14079abf94 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java @@ -19,7 +19,7 @@ package org.apache.paimon.table.sink; import org.apache.paimon.data.BinaryRow; -import org.apache.paimon.manifest.ManifestEntry; +import org.apache.paimon.manifest.PartitionEntry; import org.apache.paimon.table.FileStoreTable; import java.io.Serializable; @@ -79,9 +79,11 @@ public PartitionBucketMapping( *

    For non-partitioned tables, this returns a mapping with only the schema-defined default * bucket count and an empty partition map. * - *

    For partitioned tables, the method scans all manifest entries and records the {@code - * totalBuckets} value for each partition. If the scan fails for any reason, a fallback mapping - * with only the default bucket count is returned. + *

    For partitioned tables, the method reads {@link + * org.apache.paimon.manifest.PartitionEntry}s, which aggregate manifest entries per partition + * during the scan and therefore have a much smaller memory footprint than loading all data file + * entries. If the scan fails for any reason, a fallback mapping with only the default bucket + * count is returned. * * @param table the {@link FileStoreTable} to load the mapping from * @return a {@link PartitionBucketMapping} reflecting the current bucket layout of the table @@ -93,13 +95,15 @@ public static PartitionBucketMapping loadFromTable(FileStoreTable table) { } try { - List entries = table.store().newScan().plan().files(); + List partitionEntries = table.store().newScan().readPartitionEntries(); Map partitionBucketMap = new HashMap<>(); - for (ManifestEntry entry : entries) { + for (PartitionEntry entry : partitionEntries) { int totalBuckets = entry.totalBuckets(); - if (totalBuckets > 0) { - BinaryRow partition = entry.partition(); - partitionBucketMap.putIfAbsent(partition.copy(), totalBuckets); + // Only store partitions whose bucket count differs from the default. + // This keeps the map empty for partitions that have never been rescaled, + // avoiding per-partition BinaryRow copies and Integer allocations entirely. + if (totalBuckets > 0 && totalBuckets != defaultBuckets) { + partitionBucketMap.put(entry.partition().copy(), totalBuckets); } } From 81f5d4476bb4394538501fede22f44e2d234df26 Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Wed, 15 Apr 2026 11:46:23 +1000 Subject: [PATCH 3/7] Fix rescaling via INSERT OVERWRITE --- ...e.java => SchemaBucketFileStoreTable.java} | 20 ++--- .../paimon/operation/FileStoreCommitTest.java | 58 +++++++------- .../paimon/flink/action/CompactAction.java | 13 +--- .../paimon/flink/action/RescaleAction.java | 6 +- .../paimon/flink/sink/FlinkSinkBuilder.java | 11 ++- .../paimon/flink/RescaleBucketITCase.java | 77 +++++++++++++++++++ 6 files changed, 131 insertions(+), 54 deletions(-) rename paimon-core/src/main/java/org/apache/paimon/table/{RescaleFileStoreTable.java => SchemaBucketFileStoreTable.java} (76%) diff --git a/paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java similarity index 76% rename from paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java rename to paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java index 4d2244f8640b..88e2dfbb8894 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/RescaleFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java @@ -32,13 +32,13 @@ import java.util.Optional; /** - * A {@link FileStoreTable} wrapper for rescale operations that overrides bucket-related behavior to - * use the new target bucket count instead of loading per-partition bucket mappings from the - * manifest. + * A {@link FileStoreTable} wrapper that uses the schema number of buckets assign writes instead of + * using the number of buckets defined in each partition. Useful for postpone buckets, overrides and + * rescales. */ -public class RescaleFileStoreTable extends DelegatedFileStoreTable { +public class SchemaBucketFileStoreTable extends DelegatedFileStoreTable { - public RescaleFileStoreTable(FileStoreTable wrapped) { + public SchemaBucketFileStoreTable(FileStoreTable wrapped) { super(wrapped); } @@ -67,26 +67,26 @@ public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) @Override public FileStoreTable copy(Map dynamicOptions) { - return new RescaleFileStoreTable(wrapped().copy(dynamicOptions)); + return new SchemaBucketFileStoreTable(wrapped().copy(dynamicOptions)); } @Override public FileStoreTable copy(TableSchema newTableSchema) { - return new RescaleFileStoreTable(wrapped().copy(newTableSchema)); + return new SchemaBucketFileStoreTable(wrapped().copy(newTableSchema)); } @Override public FileStoreTable copyWithoutTimeTravel(Map dynamicOptions) { - return new RescaleFileStoreTable(wrapped().copyWithoutTimeTravel(dynamicOptions)); + return new SchemaBucketFileStoreTable(wrapped().copyWithoutTimeTravel(dynamicOptions)); } @Override public FileStoreTable copyWithLatestSchema() { - return new RescaleFileStoreTable(wrapped().copyWithLatestSchema()); + return new SchemaBucketFileStoreTable(wrapped().copyWithLatestSchema()); } @Override public FileStoreTable switchToBranch(String branchName) { - return new RescaleFileStoreTable(wrapped().switchToBranch(branchName)); + return new SchemaBucketFileStoreTable(wrapped().switchToBranch(branchName)); } } diff --git a/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java b/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java index 2efb5ab4974e..e3c31b85dc46 100644 --- a/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/operation/FileStoreCommitTest.java @@ -1091,35 +1091,35 @@ public void testBucketCountConsistencyValidation() throws Exception { // Re-commit the same data but with a different totalBuckets value. // This simulates a stale writer that loaded an old bucket mapping. assertThatThrownBy( - () -> - store.commitDataImpl( - data, - gen::getPartition, - kv -> 0, - false, - null, - null, - Collections.emptyList(), - (commit, committable) -> { - ManifestCommittable tampered = - new ManifestCommittable( - committable.identifier(), - committable.watermark()); - for (CommitMessage msg : - committable.fileCommittables()) { - CommitMessageImpl impl = (CommitMessageImpl) msg; - tampered.addFileCommittable( - new CommitMessageImpl( - impl.partition(), - impl.bucket(), - 99, - impl.newFilesIncrement(), - impl.compactIncrement())); - } - commit.commit(tampered, true); - })) - .isInstanceOf(RuntimeException.class) - .hasMessageContaining("without overwrite"); + () -> + store.commitDataImpl( + data, + gen::getPartition, + kv -> 0, + false, + null, + null, + Collections.emptyList(), + (commit, committable) -> { + ManifestCommittable tampered = + new ManifestCommittable( + committable.identifier(), + committable.watermark()); + for (CommitMessage msg : + committable.fileCommittables()) { + CommitMessageImpl impl = (CommitMessageImpl) msg; + tampered.addFileCommittable( + new CommitMessageImpl( + impl.partition(), + impl.bucket(), + 99, + impl.newFilesIncrement(), + impl.compactIncrement())); + } + commit.commit(tampered, true); + })) + .isInstanceOf(RuntimeException.class) + .hasMessageContaining("without overwrite"); } private FileStoreCommitImpl newCommitWithSnapshotCommit( diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java index 63003cffafbc..4061336fe63a 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/CompactAction.java @@ -45,8 +45,7 @@ import org.apache.paimon.predicate.PredicateProjectionConverter; import org.apache.paimon.table.BucketMode; import org.apache.paimon.table.FileStoreTable; -import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; -import org.apache.paimon.table.sink.PartitionBucketMapping; +import org.apache.paimon.table.SchemaBucketFileStoreTable; import org.apache.paimon.types.RowType; import org.apache.paimon.utils.InternalRowPartitionComputer; import org.apache.paimon.utils.Pair; @@ -337,7 +336,8 @@ protected boolean buildForPostponeBucketCompaction( bucketOptions = new HashMap<>(table.options()); bucketOptions.put(CoreOptions.BUCKET.key(), String.valueOf(partitionBucketNum)); - FileStoreTable realTable = table.copy(table.schema().copy(bucketOptions)); + FileStoreTable realTable = + new SchemaBucketFileStoreTable(table.copy(table.schema().copy(bucketOptions))); LinkedHashMap partitionSpec = partitionComputer.generatePartValues(partition); @@ -348,18 +348,13 @@ protected boolean buildForPostponeBucketCompaction( partitionSpec, options.get(FlinkConnectorOptions.SCAN_PARALLELISM)); - PartitionBucketMapping partitionBucketMapping = - new PartitionBucketMapping(partitionBucketNum); - FixedBucketRowKeyExtractor extractor = - new FixedBucketRowKeyExtractor(realTable.schema(), partitionBucketMapping); - DataStream partitioned = FlinkStreamPartitioner.partition( FlinkSinkBuilder.mapToInternalRow( sourcePair.getLeft(), realTable.rowType(), table.catalogEnvironment().catalogContext()), - new RowDataChannelComputer(extractor), + new RowDataChannelComputer(realTable.createRowKeyExtractor()), null); FixedBucketSink sink = new FixedBucketSink(realTable, null); DataStream written = diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java index 976bfb302201..3855a5115d43 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/action/RescaleAction.java @@ -26,7 +26,6 @@ import org.apache.paimon.partition.PartitionPredicate; import org.apache.paimon.table.BucketMode; import org.apache.paimon.table.FileStoreTable; -import org.apache.paimon.table.RescaleFileStoreTable; import org.apache.paimon.utils.Preconditions; import org.apache.flink.api.common.RuntimeExecutionMode; @@ -121,9 +120,8 @@ public void build() throws Exception { } else { bucketOptions.put(CoreOptions.BUCKET.key(), String.valueOf(bucketNum)); } - RescaleFileStoreTable rescaledTable = - new RescaleFileStoreTable( - fileStoreTable.copy(fileStoreTable.schema().copy(bucketOptions))); + FileStoreTable rescaledTable = + fileStoreTable.copy(fileStoreTable.schema().copy(bucketOptions)); new FlinkSinkBuilder(rescaledTable) .overwrite(partition) .parallelism(sinkParallelism == null ? bucketNum : sinkParallelism) diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java index 430eeb180c9f..aa89884c8039 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java @@ -36,6 +36,7 @@ import org.apache.paimon.table.BucketMode; import org.apache.paimon.table.FileStoreTable; import org.apache.paimon.table.PostponeUtils; +import org.apache.paimon.table.SchemaBucketFileStoreTable; import org.apache.paimon.table.Table; import org.apache.paimon.table.sink.ChannelComputer; import org.apache.paimon.utils.BlobDescriptorUtils; @@ -300,10 +301,16 @@ protected DataStreamSink buildForFixedBucket(DataStream input) { + " then the parallelism of writerOperator will be set to bucketNums."); parallelism = bucketNums; } + // When overwriting a specific partition on a partitioned fixed-bucket table, wrap the + // table in SchemaBucketFileStoreTable so that createRowKeyExtractor() uses the new schema + // bucket count for row routing, instead of loading the old per-partition bucket + // mapping from the manifest. + FileStoreTable sinkTable = + overwritePartition != null ? new SchemaBucketFileStoreTable(table) : table; RowDataChannelComputer channelComputer = - new RowDataChannelComputer(table.createRowKeyExtractor()); + new RowDataChannelComputer(sinkTable.createRowKeyExtractor()); DataStream partitioned = partition(input, channelComputer, parallelism); - FixedBucketSink sink = new FixedBucketSink(table, overwritePartition); + FixedBucketSink sink = new FixedBucketSink(sinkTable, overwritePartition); return sink.sinkFrom(partitioned); } diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java index 9b981102ec56..05ff4332ee5a 100644 --- a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java @@ -20,8 +20,10 @@ import org.apache.paimon.Snapshot; import org.apache.paimon.fs.local.LocalFileIO; +import org.apache.paimon.manifest.ManifestEntry; import org.apache.paimon.schema.SchemaManager; import org.apache.paimon.schema.TableSchema; +import org.apache.paimon.table.FileStoreTable; import org.apache.paimon.utils.SnapshotManager; import org.apache.flink.core.execution.JobClient; @@ -33,6 +35,8 @@ import javax.annotation.Nullable; import java.util.Arrays; +import java.util.Collections; +import java.util.Iterator; import java.util.List; import java.util.concurrent.CompletableFuture; import java.util.concurrent.ExecutionException; @@ -229,6 +233,79 @@ private void innerTest(String catalogName, String tableName) { .containsExactlyInAnyOrderElementsOf(expected); } + @Test + public void testRescaleSinglePartitionViaInsertOverwrite() throws Exception { + // Create a partitioned primary-key table with 2 buckets. + // Primary-key tables are sensitive to bucket routing: each row must land in the + // bucket determined by hash(pk) % totalBuckets, otherwise merging during reads + // produces duplicates (LSM finds the same key in different buckets and returns both). + batchSql("USE CATALOG fs_catalog"); + batchSql("DROP TABLE IF EXISTS `TP`"); + batchSql( + "CREATE TABLE `TP` " + + "(dt STRING, f0 INT, PRIMARY KEY (dt, f0) NOT ENFORCED) " + + "PARTITIONED BY (dt) " + + "WITH ('bucket' = '2')"); + + // Insert data into two partitions + batchSql("INSERT INTO TP VALUES ('p1', 1), ('p1', 2), ('p1', 3)"); + batchSql("INSERT INTO TP VALUES ('p2', 4), ('p2', 5), ('p2', 6)"); + + // Alter bucket count to 4 + batchSql(alterTableSql, "TP", 4); + + // INSERT OVERWRITE only partition p1 via the SQL path. + // The SQL path goes through FlinkTableSinkBase which does NOT wrap the table in + // OverwriteFileStoreTable. As a result, table.createRowKeyExtractor() calls + // PartitionBucketMapping.loadFromTable() which reads the old per-partition bucket + // count from the manifest (2 for p1) and uses it for row routing (hashing). + // Rows land in buckets [0,1] but files are stamped totalBuckets=4. + // A subsequent upsert into p1 routes rows using the new 4-bucket mapping — the + // same key may appear in two different files with different bucket assignments, + // causing duplicate reads. + batchSql("INSERT OVERWRITE TP PARTITION (dt = 'p1') SELECT f0 FROM TP WHERE dt = 'p1'"); + + // Verify the overwrite preserved data correctly + assertThat(batchSql("SELECT * FROM TP WHERE dt = 'p1'")) + .containsExactlyInAnyOrder(Row.of("p1", 1), Row.of("p1", 2), Row.of("p1", 3)); + assertThat(batchSql("SELECT * FROM TP WHERE dt = 'p2'")) + .containsExactlyInAnyOrder(Row.of("p2", 4), Row.of("p2", 5), Row.of("p2", 6)); + + // Now upsert (update) existing rows in p1 — this uses the new 4-bucket mapping. + // If the overwrite used the old 2-bucket hashing, the original row for key (p1,1) + // lands in bucket hash(1)%2, but the update lands in bucket hash(1)%4. + // The LSM reader finds two files for key (p1,1) in different buckets and returns + // both — causing duplicate rows. + batchSql("INSERT INTO TP VALUES ('p1', 1), ('p1', 2), ('p1', 3)"); + assertThat(batchSql("SELECT * FROM TP WHERE dt = 'p1'")) + .as( + "After rescale overwrite + upsert of same keys, each key must appear " + + "exactly once. If INSERT OVERWRITE used the old 2-bucket " + + "routing instead of the new 4-bucket routing (via " + + "OverwriteFileStoreTable), the same key will exist in two " + + "different buckets causing duplicate rows on read.") + .containsExactlyInAnyOrder(Row.of("p1", 1), Row.of("p1", 2), Row.of("p1", 3)); + + // Also verify the files written by INSERT OVERWRITE are stamped with the new bucket count + FileStoreTable fileStoreTable = paimonTable("TP"); + Iterator it = + fileStoreTable + .newSnapshotReader() + .withPartitionFilter(Collections.singletonMap("dt", "p1")) + .onlyReadRealBuckets() + .readFileIterator(); + assertThat(it.hasNext()).isTrue(); + while (it.hasNext()) { + ManifestEntry entry = it.next(); + assertThat(entry.totalBuckets()) + .as("Files in partition p1 must be stamped with the new bucket count (4)") + .isEqualTo(4); + assertThat(entry.bucket()).as("Bucket index must be in range [0, 3]").isBetween(0, 3); + } + + batchSql("USE CATALOG default_catalog"); + } + private void executeBoth(List sqlList) { sqlList.forEach( sql -> { From 012e72adcbf89c6e1ed29e578cd6e0865656f37a Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Thu, 14 May 2026 15:35:29 +1000 Subject: [PATCH 4/7] Fix empty-bucket from WriteRestore scenario --- .../operation/FileSystemWriteRestore.java | 13 +- .../apache/paimon/operation/WriteRestore.java | 6 +- .../table/sink/PartitionBucketMapping.java | 6 +- .../operation/FileSystemWriteRestoreTest.java | 255 ++++++++++++++++++ .../paimon/flink/sink/FlinkSinkBuilder.java | 6 +- .../PostponeFixedBucketChannelComputer.java | 9 +- .../coordinator/TableWriteCoordinator.java | 19 +- .../paimon/flink/RescaleBucketITCase.java | 155 +++++++++++ ...ostponeFixedBucketChannelComputerTest.java | 13 +- .../TableWriteCoordinatorTest.java | 105 ++++++++ 10 files changed, 572 insertions(+), 15 deletions(-) create mode 100644 paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java diff --git a/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java b/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java index e7faf2a24569..9463b014ba31 100644 --- a/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java +++ b/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java @@ -25,9 +25,9 @@ import org.apache.paimon.index.IndexFileMeta; import org.apache.paimon.io.DataFileMeta; import org.apache.paimon.manifest.ManifestEntry; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.utils.SnapshotManager; -import java.util.ArrayList; import java.util.List; import static org.apache.paimon.deletionvectors.DeletionVectorsIndexFile.DELETION_VECTORS_INDEX; @@ -38,6 +38,7 @@ public class FileSystemWriteRestore implements WriteRestore { private final SnapshotManager snapshotManager; private final FileStoreScan scan; private final IndexFileHandler indexFileHandler; + private final PartitionBucketMapping partitionBucketMapping; public FileSystemWriteRestore( CoreOptions options, @@ -52,6 +53,7 @@ public FileSystemWriteRestore( this.scan.dropStats(); } } + this.partitionBucketMapping = PartitionBucketMapping.loadFromScan(scan, options.bucket()); } @Override @@ -75,10 +77,15 @@ public RestoreFiles restoreFiles( return RestoreFiles.empty(); } - List restoreFiles = new ArrayList<>(); List entries = scan.withSnapshot(snapshot).withPartitionBucket(partition, bucket).plan().files(); - Integer totalBuckets = WriteRestore.extractDataFiles(entries, restoreFiles); + List restoreFiles = WriteRestore.extractDataFiles(entries); + + // Resolve the totalBuckets from the partition-level mapping rather than + // using the bucket entries, because a partition may have been rescaled to a + // different totalBuckets. If we used the bucket entries for an empty bucket, + // we would incorrectly return the table default totalBuckets for that partition. + Integer totalBuckets = partitionBucketMapping.resolveNumBuckets(partition); IndexFileMeta dynamicBucketIndex = null; if (scanDynamicBucketIndex) { diff --git a/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java b/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java index 5d4e335571d1..987f4f04b85c 100644 --- a/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java +++ b/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java @@ -24,6 +24,7 @@ import javax.annotation.Nullable; +import java.util.ArrayList; import java.util.List; /** Restore for write to restore data files by partition and bucket from file system. */ @@ -38,8 +39,9 @@ RestoreFiles restoreFiles( boolean scanDeleteVectorsIndex); @Nullable - static Integer extractDataFiles(List entries, List dataFiles) { + static List extractDataFiles(List entries) { Integer totalBuckets = null; + List dataFiles = new ArrayList<>(); for (ManifestEntry entry : entries) { if (totalBuckets != null && totalBuckets != entry.totalBuckets()) { throw new RuntimeException( @@ -50,6 +52,6 @@ static Integer extractDataFiles(List entries, List totalBuckets = entry.totalBuckets(); dataFiles.add(entry.file()); } - return totalBuckets; + return dataFiles; } } diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java index bf14079abf94..a1d98c24af7e 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java @@ -20,6 +20,7 @@ import org.apache.paimon.data.BinaryRow; import org.apache.paimon.manifest.PartitionEntry; +import org.apache.paimon.operation.FileStoreScan; import org.apache.paimon.table.FileStoreTable; import java.io.Serializable; @@ -93,9 +94,12 @@ public static PartitionBucketMapping loadFromTable(FileStoreTable table) { if (table.partitionKeys().isEmpty()) { return new PartitionBucketMapping(defaultBuckets, Collections.emptyMap()); } + return loadFromScan(table.store().newScan(), defaultBuckets); + } + public static PartitionBucketMapping loadFromScan(FileStoreScan scan, int defaultBuckets) { try { - List partitionEntries = table.store().newScan().readPartitionEntries(); + List partitionEntries = scan.readPartitionEntries(); Map partitionBucketMap = new HashMap<>(); for (PartitionEntry entry : partitionEntries) { int totalBuckets = entry.totalBuckets(); diff --git a/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java b/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java new file mode 100644 index 000000000000..1470f92a3747 --- /dev/null +++ b/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java @@ -0,0 +1,255 @@ +/* + * 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.paimon.operation; + +import org.apache.paimon.CoreOptions; +import org.apache.paimon.catalog.Identifier; +import org.apache.paimon.data.BinaryRow; +import org.apache.paimon.data.BinaryRowWriter; +import org.apache.paimon.data.GenericRow; +import org.apache.paimon.schema.Schema; +import org.apache.paimon.table.FileStoreTable; +import org.apache.paimon.table.TableTestBase; +import org.apache.paimon.table.sink.CommitMessage; +import org.apache.paimon.table.sink.CommitMessageImpl; +import org.apache.paimon.table.sink.StreamTableWrite; +import org.apache.paimon.table.sink.StreamWriteBuilder; +import org.apache.paimon.table.sink.TableCommitImpl; +import org.apache.paimon.types.DataTypes; + +import org.junit.jupiter.api.Test; + +import java.util.Collections; +import java.util.List; +import java.util.UUID; + +import static org.assertj.core.api.Assertions.assertThat; + +/** + * Tests for {@link FileSystemWriteRestore} covering the scenario where a partition has a different + * bucket count than the table default (e.g., after a rescale operation). + */ +public class FileSystemWriteRestoreTest extends TableTestBase { + + /** + * Scenario: + * + *

      + *
    • Table bucket (default): 32 + *
    • Partition A bucket: 2 (rescaled) + *
    • Partition A bucket=0: has data files, totalBuckets=2 + *
    • Partition A bucket=1: no data files + *
    + * + *

    When restoring bucket=1 of partition A (empty bucket), the returned {@code totalBuckets} + * must be 2 (from the per-partition mapping), not 32 (the table default). + */ + @Test + public void testEmptyBucketUsesPartitionBucketCount() throws Exception { + // Table default: 32 buckets + int tableBuckets = 32; + // Partition A was rescaled to 2 buckets + int partitionBuckets = 2; + + Identifier identifier = new Identifier("db", "table"); + catalog.createDatabase("db", false); + Schema schema = + Schema.newBuilder() + .column("pt", DataTypes.INT()) + .column("k", DataTypes.INT()) + .column("v", DataTypes.INT()) + .primaryKey("pt", "k") + .partitionKeys("pt") + .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) + .build(); + catalog.createTable(identifier, schema, false); + FileStoreTable table = getTable(identifier); + + // Write data to partition A, bucket=0 with totalBuckets=2 (simulating a rescaled + // partition). We write normally first, then re-wrap the commit message to override + // totalBuckets. + BinaryRow partitionA = partitionRow(1); + String commitUser = UUID.randomUUID().toString(); + StreamWriteBuilder writeBuilder = table.newStreamWriteBuilder().withCommitUser(commitUser); + List messages; + try (StreamTableWrite write = writeBuilder.newWrite()) { + // Write to partition=1, key=1, value=100 with bucket=0 explicitly + write.write(GenericRow.of(1, 1, 100), 0); + messages = write.prepareCommit(false, 0); + } + + // Rewrap the commit message so that totalBuckets=2 (the rescaled partition bucket count) + CommitMessageImpl original = (CommitMessageImpl) messages.get(0); + CommitMessageImpl rescaled = + new CommitMessageImpl( + original.partition(), + original.bucket(), + partitionBuckets, + original.newFilesIncrement(), + original.compactIncrement()); + + try (TableCommitImpl commit = table.newCommit(commitUser)) { + commit.commit(0, Collections.singletonList(rescaled)); + } + + // Now create the FileSystemWriteRestore for this table + FileStoreTable freshTable = getTable(identifier); + FileSystemWriteRestore writeRestore = newWriteRestore(freshTable); + + // Restore bucket=0 (has data files): totalBuckets should be 2 + RestoreFiles restored0 = writeRestore.restoreFiles(partitionA, 0, false, false); + assertThat(restored0.totalBuckets()) + .as("bucket=0 (has files) should use partition bucket count, not table default") + .isEqualTo(partitionBuckets); + assertThat(restored0.dataFiles()).isNotEmpty(); + + // Restore bucket=1 (empty bucket): totalBuckets should ALSO be 2, not 32 + RestoreFiles restored1 = writeRestore.restoreFiles(partitionA, 1, false, false); + assertThat(restored1.totalBuckets()) + .as("bucket=1 (empty bucket) should use partition bucket count, not table default") + .isEqualTo(partitionBuckets); + assertThat(restored1.dataFiles()).isEmpty(); + } + + /** + * Sanity check: a partition that has never been rescaled uses the table default bucket count. + */ + @Test + public void testPartitionWithDefaultBucketCount() throws Exception { + int tableBuckets = 32; + + Identifier identifier = new Identifier("db2", "table"); + catalog.createDatabase("db2", false); + Schema schema = + Schema.newBuilder() + .column("pt", DataTypes.INT()) + .column("k", DataTypes.INT()) + .column("v", DataTypes.INT()) + .primaryKey("pt", "k") + .partitionKeys("pt") + .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) + .build(); + catalog.createTable(identifier, schema, false); + FileStoreTable table = getTable(identifier); + + // Write data using the default bucket count (no totalBuckets override) + write(table, GenericRow.of(1, 1, 100)); + + FileStoreTable freshTable = getTable(identifier); + FileSystemWriteRestore writeRestore = newWriteRestore(freshTable); + + BinaryRow partitionA = partitionRow(1); + + // Restore bucket=0 (has data with default totalBuckets=32) + RestoreFiles restored = writeRestore.restoreFiles(partitionA, 0, false, false); + assertThat(restored.totalBuckets()) + .as("partition with default bucket count should return table bucket count") + .isEqualTo(tableBuckets); + } + + /** + * Scenario with two partitions: partition A rescaled to 2 buckets, partition B uses default 32. + * Each partition's empty buckets must return their own bucket count. + */ + @Test + public void testMixedPartitionsWithDifferentBucketCounts() throws Exception { + int tableBuckets = 32; + int partitionABuckets = 2; + + Identifier identifier = new Identifier("db3", "table"); + catalog.createDatabase("db3", false); + Schema schema = + Schema.newBuilder() + .column("pt", DataTypes.INT()) + .column("k", DataTypes.INT()) + .column("v", DataTypes.INT()) + .primaryKey("pt", "k") + .partitionKeys("pt") + .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) + .build(); + catalog.createTable(identifier, schema, false); + FileStoreTable table = getTable(identifier); + + String commitUser = UUID.randomUUID().toString(); + + // Write partition A, bucket=0 with rescaled totalBuckets=2 + StreamWriteBuilder writeBuilder = table.newStreamWriteBuilder().withCommitUser(commitUser); + List messagesA; + try (StreamTableWrite write = writeBuilder.newWrite()) { + write.write(GenericRow.of(1, 1, 100), 0); + messagesA = write.prepareCommit(false, 0); + } + CommitMessageImpl originalA = (CommitMessageImpl) messagesA.get(0); + CommitMessageImpl rescaledA = + new CommitMessageImpl( + originalA.partition(), + originalA.bucket(), + partitionABuckets, + originalA.newFilesIncrement(), + originalA.compactIncrement()); + try (TableCommitImpl commit = table.newCommit(commitUser)) { + commit.commit(0, Collections.singletonList(rescaledA)); + } + + // Write partition B, bucket=0 with default totalBuckets=32 + try (StreamTableWrite write = writeBuilder.newWrite()) { + write.write(GenericRow.of(2, 1, 200), 0); + List messagesB = write.prepareCommit(false, 1); + try (TableCommitImpl commit = table.newCommit(commitUser)) { + commit.commit(1, messagesB); + } + } + + FileStoreTable freshTable = getTable(identifier); + FileSystemWriteRestore writeRestore = newWriteRestore(freshTable); + + BinaryRow partitionA = partitionRow(1); + BinaryRow partitionB = partitionRow(2); + + // Partition A: bucket=1 (empty) should use 2, not 32 + RestoreFiles restoredA1 = writeRestore.restoreFiles(partitionA, 1, false, false); + assertThat(restoredA1.totalBuckets()) + .as("partition A empty bucket should use rescaled partition bucket count 2") + .isEqualTo(partitionABuckets); + assertThat(restoredA1.dataFiles()).isEmpty(); + + // Partition B: bucket=0 (has files) should use 32 + RestoreFiles restoredB0 = writeRestore.restoreFiles(partitionB, 0, false, false); + assertThat(restoredB0.totalBuckets()) + .as("partition B should use table default bucket count 32") + .isEqualTo(tableBuckets); + assertThat(restoredB0.dataFiles()).isNotEmpty(); + } + + private static FileSystemWriteRestore newWriteRestore(FileStoreTable table) { + return new FileSystemWriteRestore( + table.store().options(), + table.store().snapshotManager(), + table.store().newScan(), + table.store().newIndexFileHandler()); + } + + private BinaryRow partitionRow(int partitionValue) { + BinaryRow row = new BinaryRow(1); + BinaryRowWriter writer = new BinaryRowWriter(row); + writer.writeInt(0, partitionValue); + writer.complete(); + return row; + } +} diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java index aa89884c8039..d79751248f8f 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java @@ -39,6 +39,7 @@ import org.apache.paimon.table.SchemaBucketFileStoreTable; import org.apache.paimon.table.Table; import org.apache.paimon.table.sink.ChannelComputer; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.utils.BlobDescriptorUtils; import org.apache.flink.api.common.functions.MapFunction; @@ -328,10 +329,13 @@ private DataStreamSink buildPostponeBucketSink(DataStream input) return sink.sinkFrom(partitioned); } else { Map knownNumBuckets = PostponeUtils.getKnownNumBuckets(table); + PartitionBucketMapping partitionBucketMapping = + PartitionBucketMapping.loadFromTable(table); DataStream partitioned = partition( input, - new PostponeFixedBucketChannelComputer(table.schema(), knownNumBuckets), + new PostponeFixedBucketChannelComputer( + table.schema(), knownNumBuckets, partitionBucketMapping), parallelism); FileStoreTable tableForWrite = PostponeUtils.tableForFixBucketWrite(table); diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputer.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputer.java index e1bf08d8434c..98cb9a07ee83 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputer.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputer.java @@ -23,6 +23,7 @@ import org.apache.paimon.schema.TableSchema; import org.apache.paimon.table.sink.ChannelComputer; import org.apache.paimon.table.sink.FixedBucketRowKeyExtractor; +import org.apache.paimon.table.sink.PartitionBucketMapping; import java.util.Map; @@ -36,20 +37,24 @@ public class PostponeFixedBucketChannelComputer implements ChannelComputer knownNumBuckets; + private final PartitionBucketMapping partitionBucketMapping; private transient int numChannels; private transient FixedBucketRowKeyExtractor keyExtractor; public PostponeFixedBucketChannelComputer( - TableSchema schema, Map knownNumBuckets) { + TableSchema schema, + Map knownNumBuckets, + PartitionBucketMapping partitionBucketMapping) { this.schema = schema; this.knownNumBuckets = knownNumBuckets; + this.partitionBucketMapping = partitionBucketMapping; } @Override public void setup(int numChannels) { this.numChannels = numChannels; - this.keyExtractor = new FixedBucketRowKeyExtractor(schema); + this.keyExtractor = new FixedBucketRowKeyExtractor(schema, partitionBucketMapping); } @Override diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java index 0f315880e932..6d3dea1eddef 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java @@ -28,13 +28,13 @@ import org.apache.paimon.operation.FileStoreScan; import org.apache.paimon.operation.WriteRestore; import org.apache.paimon.table.FileStoreTable; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.shade.caffeine2.com.github.benmanes.caffeine.cache.Cache; import org.apache.paimon.shade.caffeine2.com.github.benmanes.caffeine.cache.Caffeine; import java.io.IOException; import java.time.Duration; -import java.util.ArrayList; import java.util.Arrays; import java.util.List; import java.util.Map; @@ -62,6 +62,7 @@ public class TableWriteCoordinator { private final Cache pagedCoordination; private volatile Snapshot snapshot; + private volatile PartitionBucketMapping partitionBucketMapping; public TableWriteCoordinator(FileStoreTable table) { this.table = table; @@ -93,6 +94,7 @@ private synchronized void refresh() { } this.snapshot = latestSnapshot.get(); this.scan.withSnapshot(snapshot); + loadPartitionBucketMapping(); } public synchronized PagedCoordinationResponse scan(PagedCoordinationRequest request) @@ -148,9 +150,14 @@ public synchronized ScanCoordinationResponse scan(ScanCoordinationRequest reques BinaryRow partition = deserializeBinaryRow(request.partition()); int bucket = request.bucket(); - List restoreFiles = new ArrayList<>(); List entries = scan.withPartitionBucket(partition, bucket).plan().files(); - Integer totalBuckets = WriteRestore.extractDataFiles(entries, restoreFiles); + List restoreFiles = WriteRestore.extractDataFiles(entries); + + // Resolve the totalBuckets from the partition-level mapping rather than + // using the bucket entries, because a partition may have been rescaled to a + // different totalBuckets. If we used the bucket entries for an empty bucket, + // we would incorrectly return the table default totalBuckets for that partition. + Integer totalBuckets = partitionBucketMapping.resolveNumBuckets(partition); IndexFileMeta dynamicBucketIndex = null; if (request.scanDynamicBucketIndex()) { @@ -182,6 +189,7 @@ private synchronized long computeLatestIdentifier(String user) { if (snapshot == null || latestSnapshotOfUser.id() > snapshot.id()) { snapshot = latestSnapshotOfUser; scan.withSnapshot(snapshot); + loadPartitionBucketMapping(); } return latestSnapshotOfUser.commitIdentifier(); } @@ -193,6 +201,11 @@ public void checkpoint() { latestCommittedIdentifiers.clear(); } + private void loadPartitionBucketMapping() { + int defaultNumBuckets = table.schema().numBuckets(); + this.partitionBucketMapping = PartitionBucketMapping.loadFromScan(scan, defaultNumBuckets); + } + private static class CoordinationKey { private final byte[] content; diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java index 05ff4332ee5a..e33eef3fef3c 100644 --- a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/RescaleBucketITCase.java @@ -306,6 +306,161 @@ public void testRescaleSinglePartitionViaInsertOverwrite() throws Exception { batchSql("USE CATALOG default_catalog"); } + @Test + public void testWriteToEmptyBucketAfterRescaleKeepsPartitionBucketCount() throws Exception { + // End-to-end reproduction of the FileSystemWriteRestore empty-bucket scenario. + // + // The scenario occurs when a partition has a bucket count that differs from + // the table-level default, AND at least one of that partition's buckets + // has NO existing data files. A subsequent write that lands in the + // empty bucket must keep the partition's bucket count, not silently + // adopt the table default. + // + // Final state we want before the bug-triggering write (Step 5): + // - table default bucket : 2 + // - partition p1 bucket : 4 (held over from a prior rescale+overwrite) + // - partition p1 bucket=X: has data files, totalBuckets = 4 + // - partition p1 bucket=Y: NO data files + // + // The Step 5 write fills bucket=Y in p1. Post-fix, the new file MUST + // be stamped totalBuckets=4. Before the fix, FileSystemWriteRestore + // returned null for the empty bucket, the writer fell back to the + // table default (2), and the new file was stamped totalBuckets=2, + // corrupting p1's bucket layout. + + // Step 1: create a partitioned PK table with bucket=2 and seed one row + // into p1. + batchSql( + "CREATE TABLE IF NOT EXISTS `fs_catalog`.`default`.`TPE` " + + "(dt STRING, f0 INT, PRIMARY KEY (dt, f0) NOT ENFORCED) " + + "PARTITIONED BY (dt) " + + "WITH ('bucket' = '2')"); + batchSql("USE CATALOG fs_catalog"); + batchSql("INSERT INTO TPE VALUES ('p1', 1)"); + // Also seed partition p2 — this acts as a positive control: p2 is + // never rescaled and must always be stamped with the current table + // default bucket count. + batchSql("INSERT INTO TPE VALUES ('p2', 1)"); + + // Step 2: rescale the table default to 4. + batchSql(alterTableSql, "TPE", 4); + + // Step 3: INSERT OVERWRITE p1 — this rewrites p1's files using the + // current table default (4), so afterwards p1's files are stamped + // totalBuckets=4. With only a single row in p1, at least one of p1's + // four buckets is guaranteed to be empty. + batchSql("INSERT OVERWRITE TPE PARTITION (dt = 'p1') SELECT f0 FROM TPE WHERE dt = 'p1'"); + + // Step 4 (NEW): rescale the table default BACK to 2, while p1 keeps + // its 4-bucket layout. Now the per-partition bucket count (4) diverges + // from the table default (2) — the precondition for the bug. + batchSql(alterTableSql, "TPE", 2); + + // Step 5: insert more rows into p1. PartitionBucketMapping reads the + // manifest and routes these rows using p1's actual bucket count (4). + // Several rows are inserted so that — with high probability — at + // least one lands in the bucket that was previously empty, exercising + // the FileSystemWriteRestore empty-bucket code path. + // Also insert another row into p2 — p2 has never been rescaled and + // must remain stamped with the current table default bucket count (2). + batchSql( + "INSERT INTO TPE VALUES " + + "('p1', 2), ('p1', 3), ('p1', 4), ('p1', 5), ('p1', 6), " + + "('p2', 2), ('p2', 3), ('p2', 4), ('p2', 5), ('p2', 6)"); + + // Sanity check: each PK appears exactly once. Duplicates would indicate + // that the new rows landed in an inconsistent bucket layout (e.g. some + // files stamped totalBuckets=2 and others stamped totalBuckets=4), + // causing the LSM reader to find the same PK in two different buckets. + assertThat(batchSql("SELECT * FROM TPE")) + .as( + "Each PK must appear exactly once across all partitions. Duplicates " + + "indicate rows landed in inconsistent bucket layouts.") + .containsExactlyInAnyOrder( + Row.of("p1", 1), + Row.of("p1", 2), + Row.of("p1", 3), + Row.of("p1", 4), + Row.of("p1", 5), + Row.of("p1", 6), + Row.of("p2", 1), + Row.of("p2", 2), + Row.of("p2", 3), + Row.of("p2", 4), + Row.of("p2", 5), + Row.of("p2", 6)); + + // Strong assertion (p1 — the rescaled partition): every file must be + // stamped totalBuckets=4 (the partition's actual bucket count), NOT 2 + // (the new table default). This is the precise condition the fix + // guarantees for partitions whose bucket count differs from the table + // default. + FileStoreTable fileStoreTable = paimonTable("TPE"); + Iterator p1Iter = + fileStoreTable + .newSnapshotReader() + .withPartitionFilter(Collections.singletonMap("dt", "p1")) + .onlyReadRealBuckets() + .readFileIterator(); + assertThat(p1Iter.hasNext()).isTrue(); + while (p1Iter.hasNext()) { + ManifestEntry entry = p1Iter.next(); + assertThat(entry.totalBuckets()) + .as( + "Files in partition p1 must keep the partition's bucket count (4). " + + "If 2, FileSystemWriteRestore failed to fall back to " + + "PartitionBucketMapping for buckets that had no existing " + + "files at restore time, and stamped new files with the " + + "table-level default bucket count instead.") + .isEqualTo(4); + assertThat(entry.bucket()).as("Bucket index must be in range [0, 3]").isBetween(0, 3); + } + + // Positive control (p2 — never rescaled): every file must be stamped + // with the current table default (2). This guards against an + // over-eager fix that would erroneously route ALL writes through + // PartitionBucketMapping even when no per-partition override exists. + Iterator p2Iter = + fileStoreTable + .newSnapshotReader() + .withPartitionFilter(Collections.singletonMap("dt", "p2")) + .onlyReadRealBuckets() + .readFileIterator(); + assertThat(p2Iter.hasNext()).isTrue(); + java.util.List p2Entries = new java.util.ArrayList<>(); + while (p2Iter.hasNext()) { + p2Entries.add(p2Iter.next()); + } + // Diagnostic: dump every manifest entry in p2 so we can see what each + // file is stamped with (FileKind, bucket, totalBuckets, file name, + // snapshot/sequence info via the file meta). + System.out.println("=== p2 manifest entries (" + p2Entries.size() + " total) ==="); + for (int i = 0; i < p2Entries.size(); i++) { + ManifestEntry entry = p2Entries.get(i); + System.out.printf( + " [%d] kind=%s level=%d bucket=%d totalBuckets=%d file=%s " + + "minSeq=%d maxSeq=%d rowCount=%d%n", + i, + entry.kind(), + entry.level(), + entry.bucket(), + entry.totalBuckets(), + entry.file().fileName(), + entry.file().minSequenceNumber(), + entry.file().maxSequenceNumber(), + entry.file().rowCount()); + } + for (ManifestEntry entry : p2Entries) { + assertThat(entry.totalBuckets()) + .as( + "Files in partition p2 must use the current table default bucket " + + "count (2). p2 was never rescaled, so its files should " + + "always reflect the table-level default.") + .isEqualTo(2); + assertThat(entry.bucket()).as("Bucket index must be in range [0, 1]").isBetween(0, 1); + } + } + private void executeBoth(List sqlList) { sqlList.forEach( sql -> { diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputerTest.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputerTest.java index 991ab5d2ddbd..e2d1ffc27719 100644 --- a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputerTest.java +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/PostponeFixedBucketChannelComputerTest.java @@ -26,6 +26,7 @@ import org.apache.paimon.schema.Schema; import org.apache.paimon.schema.SchemaManager; import org.apache.paimon.schema.TableSchema; +import org.apache.paimon.table.sink.PartitionBucketMapping; import org.apache.paimon.types.DataType; import org.apache.paimon.types.DataTypes; import org.apache.paimon.types.RowType; @@ -72,8 +73,10 @@ public void testRecordsDistributedAcrossChannels() throws Exception { int numChannels = 8; Map knownNumBuckets = new HashMap<>(); + PartitionBucketMapping bucketMapping = + new PartitionBucketMapping(numChannels, Collections.emptyMap()); PostponeFixedBucketChannelComputer computer = - new PostponeFixedBucketChannelComputer(schema, knownNumBuckets); + new PostponeFixedBucketChannelComputer(schema, knownNumBuckets, bucketMapping); computer.setup(numChannels); Set channels = new HashSet<>(); @@ -113,8 +116,10 @@ public void testNoPartitionDistribution() throws Exception { int numChannels = 8; Map knownNumBuckets = new HashMap<>(); + PartitionBucketMapping bucketMapping = + new PartitionBucketMapping(numChannels, Collections.emptyMap()); PostponeFixedBucketChannelComputer computer = - new PostponeFixedBucketChannelComputer(schema, knownNumBuckets); + new PostponeFixedBucketChannelComputer(schema, knownNumBuckets, bucketMapping); computer.setup(numChannels); Set channels = new HashSet<>(); @@ -155,8 +160,10 @@ public void testSameKeyGoesToSameChannel() throws Exception { int numChannels = 8; Map knownNumBuckets = new HashMap<>(); + PartitionBucketMapping bucketMapping = + new PartitionBucketMapping(numChannels, Collections.emptyMap()); PostponeFixedBucketChannelComputer computer = - new PostponeFixedBucketChannelComputer(schema, knownNumBuckets); + new PostponeFixedBucketChannelComputer(schema, knownNumBuckets, bucketMapping); computer.setup(numChannels); // Same key should always route to the same channel diff --git a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinatorTest.java b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinatorTest.java index 8bc952f9baf4..c810e860bc5d 100644 --- a/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinatorTest.java +++ b/paimon-flink/paimon-flink-common/src/test/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinatorTest.java @@ -18,18 +18,30 @@ package org.apache.paimon.flink.sink.coordinator; +import org.apache.paimon.CoreOptions; import org.apache.paimon.Snapshot; import org.apache.paimon.catalog.Identifier; +import org.apache.paimon.data.BinaryRow; +import org.apache.paimon.data.BinaryRowWriter; import org.apache.paimon.data.GenericRow; import org.apache.paimon.schema.Schema; import org.apache.paimon.table.FileStoreTable; import org.apache.paimon.table.TableTestBase; +import org.apache.paimon.table.sink.CommitMessage; +import org.apache.paimon.table.sink.CommitMessageImpl; +import org.apache.paimon.table.sink.StreamTableWrite; +import org.apache.paimon.table.sink.StreamWriteBuilder; +import org.apache.paimon.table.sink.TableCommitImpl; import org.apache.paimon.types.DataTypes; import org.junit.jupiter.api.Test; import org.junit.jupiter.params.ParameterizedTest; import org.junit.jupiter.params.provider.ValueSource; +import java.util.Collections; +import java.util.List; +import java.util.UUID; + import static org.apache.paimon.data.BinaryRow.EMPTY_ROW; import static org.apache.paimon.utils.SerializationUtils.serializeBinaryRow; import static org.assertj.core.api.Assertions.assertThat; @@ -76,4 +88,97 @@ public void testNoManifestCache() throws Exception { assertThatThrownBy(() -> new TableWriteCoordinator(table)) .isInstanceOf(NullPointerException.class); } + + /** + * Tests that when a partition has been rescaled (different bucket count than the table + * default), the coordinator returns the correct per-partition bucket count even for buckets + * with no data files. + * + *

    Scenario: + * + *

      + *
    • Table bucket (default): 32 + *
    • Partition A bucket: 2 (rescaled) + *
    • Partition A bucket=0: has data files, totalBuckets=2 + *
    • Partition A bucket=1: no data files + *
    + * + *

    When scanning bucket=1 of partition A (empty bucket), the response {@code totalBuckets} + * must be 2 (from the per-partition mapping), not 32 (the table default). + */ + @Test + public void testEmptyBucketUsesPartitionBucketCount() throws Exception { + // Table default: 32 buckets; partition A was rescaled to 2 + int tableBuckets = 32; + int partitionBuckets = 2; + + Identifier identifier = new Identifier("db2", "table"); + catalog.createDatabase("db2", false); + Schema schema = + Schema.newBuilder() + .column("pt", DataTypes.INT()) + .column("k", DataTypes.INT()) + .column("v", DataTypes.INT()) + .primaryKey("pt", "k") + .partitionKeys("pt") + .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) + .build(); + catalog.createTable(identifier, schema, false); + FileStoreTable table = getTable(identifier); + + // Write to partition A, bucket=0 with totalBuckets=2 (rescaled partition) + String commitUser = UUID.randomUUID().toString(); + StreamWriteBuilder writeBuilder = table.newStreamWriteBuilder().withCommitUser(commitUser); + List messages; + try (StreamTableWrite write = writeBuilder.newWrite()) { + write.write(GenericRow.of(1, 1, 100), 0); + messages = write.prepareCommit(false, 0); + } + + // Override totalBuckets to simulate partition rescale to 2 + CommitMessageImpl original = (CommitMessageImpl) messages.get(0); + CommitMessageImpl rescaled = + new CommitMessageImpl( + original.partition(), + original.bucket(), + partitionBuckets, + original.newFilesIncrement(), + original.compactIncrement()); + try (TableCommitImpl commit = table.newCommit(commitUser)) { + commit.commit(0, Collections.singletonList(rescaled)); + } + + // Reload the table and create coordinator + FileStoreTable freshTable = getTable(identifier); + TableWriteCoordinator coordinator = new TableWriteCoordinator(freshTable); + + BinaryRow partitionA = partitionRow(1); + + // Scan bucket=0 (has files): totalBuckets should be 2 + ScanCoordinationRequest requestBucket0 = + new ScanCoordinationRequest(serializeBinaryRow(partitionA), 0, false, false); + ScanCoordinationResponse responseBucket0 = coordinator.scan(requestBucket0); + assertThat(responseBucket0.totalBuckets()) + .as("bucket=0 (has files) should use per-partition bucket count 2") + .isEqualTo(partitionBuckets); + assertThat(responseBucket0.extractDataFiles()).isNotEmpty(); + + // Scan bucket=1 (empty): totalBuckets must be 2, not the table default 32 + ScanCoordinationRequest requestBucket1 = + new ScanCoordinationRequest(serializeBinaryRow(partitionA), 1, false, false); + ScanCoordinationResponse responseBucket1 = coordinator.scan(requestBucket1); + assertThat(responseBucket1.totalBuckets()) + .as( + "bucket=1 (empty bucket) must use per-partition bucket count 2, not table default 32") + .isEqualTo(partitionBuckets); + assertThat(responseBucket1.extractDataFiles()).isEmpty(); + } + + private BinaryRow partitionRow(int partitionValue) { + BinaryRow row = new BinaryRow(1); + BinaryRowWriter writer = new BinaryRowWriter(row); + writer.writeInt(0, partitionValue); + writer.complete(); + return row; + } } From 0d363ce13d0b0afc6d3a21944dafe47a4351e190 Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Fri, 15 May 2026 14:24:00 +1000 Subject: [PATCH 5/7] Fixing corner case for non-partitioned tables --- .../paimon/manifest/PartitionEntry.java | 12 +- .../operation/FileSystemWriteRestore.java | 7 +- .../apache/paimon/operation/WriteRestore.java | 35 ++ .../paimon/manifest/PartitionEntryTest.java | 130 +++++++ .../operation/FileSystemWriteRestoreTest.java | 367 +++++++++--------- .../coordinator/TableWriteCoordinator.java | 8 +- 6 files changed, 354 insertions(+), 205 deletions(-) create mode 100644 paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java diff --git a/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java b/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java index d162234cd58a..43ba6d1a0f48 100644 --- a/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java +++ b/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java @@ -83,13 +83,21 @@ public int totalBuckets() { } public PartitionEntry merge(PartitionEntry entry) { + // Use the totalBuckets from the most recently created file. This correctly handles + // the case where a partition has been overwritten with a different bucket count: the + // newer files carry the new totalBuckets, and their creation time is always later. + long newLastCreationTime = Math.max(lastFileCreationTime, entry.lastFileCreationTime); + int newTotalBuckets = + lastFileCreationTime >= entry.lastFileCreationTime + ? totalBuckets + : entry.totalBuckets; return new PartitionEntry( partition, recordCount + entry.recordCount, fileSizeInBytes + entry.fileSizeInBytes, fileCount + entry.fileCount, - Math.max(lastFileCreationTime, entry.lastFileCreationTime), - entry.totalBuckets); + newLastCreationTime, + newTotalBuckets); } public Partition toPartition(InternalRowPartitionComputer computer) { diff --git a/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java b/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java index 9463b014ba31..3548123bcd68 100644 --- a/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java +++ b/paimon-core/src/main/java/org/apache/paimon/operation/FileSystemWriteRestore.java @@ -81,11 +81,8 @@ public RestoreFiles restoreFiles( scan.withSnapshot(snapshot).withPartitionBucket(partition, bucket).plan().files(); List restoreFiles = WriteRestore.extractDataFiles(entries); - // Resolve the totalBuckets from the partition-level mapping rather than - // using the bucket entries, because a partition may have been rescaled to a - // different totalBuckets. If we used the bucket entries for an empty bucket, - // we would incorrectly return the table default totalBuckets for that partition. - Integer totalBuckets = partitionBucketMapping.resolveNumBuckets(partition); + Integer totalBuckets = + WriteRestore.extractTotalBuckets(entries, partition, partitionBucketMapping); IndexFileMeta dynamicBucketIndex = null; if (scanDynamicBucketIndex) { diff --git a/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java b/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java index 987f4f04b85c..b5b963f62e7d 100644 --- a/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java +++ b/paimon-core/src/main/java/org/apache/paimon/operation/WriteRestore.java @@ -21,6 +21,7 @@ import org.apache.paimon.data.BinaryRow; import org.apache.paimon.io.DataFileMeta; import org.apache.paimon.manifest.ManifestEntry; +import org.apache.paimon.table.sink.PartitionBucketMapping; import javax.annotation.Nullable; @@ -38,7 +39,41 @@ RestoreFiles restoreFiles( boolean scanDynamicBucketIndex, boolean scanDeleteVectorsIndex); + /** + * Resolves the {@code totalBuckets} for a (partition, bucket) pair given the manifest entries + * for that bucket and the table's partition-bucket mapping. + * + *

      + *
    • Non-empty bucket: use the value stamped on the existing data files so that + * committer-side bucket-count mismatch detection (e.g. rescale-without-overwrite) still + * fires. + *
    • Empty bucket on a partitioned table: look up the per-partition override in {@code + * mapping}; returns {@code null} if the partition uses the table default. + *
    • Empty bucket on an unpartitioned table: returns {@code null} so the write path falls + * back to {@code numBuckets} and the committer-side check still fires. + *
    + */ @Nullable + static Integer extractTotalBuckets( + List entries, BinaryRow partition, PartitionBucketMapping mapping) { + if (!entries.isEmpty()) { + return entries.get(0).totalBuckets(); + } + if (partition.getFieldCount() > 0) { + return mapping.resolveNumBuckets(partition); + } + return null; + } + + /** + * Extracts the {@link DataFileMeta} list from the given manifest entries, validating that all + * entries agree on {@code totalBuckets}. + * + * @param entries manifest entries for a single (partition, bucket) pair + * @return the list of data files; empty if {@code entries} is empty + * @throws RuntimeException if entries carry inconsistent {@code totalBuckets} values, which + * indicates a corrupted manifest + */ static List extractDataFiles(List entries) { Integer totalBuckets = null; List dataFiles = new ArrayList<>(); diff --git a/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java b/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java new file mode 100644 index 000000000000..526221690a2b --- /dev/null +++ b/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java @@ -0,0 +1,130 @@ +/* + * 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.paimon.manifest; + +import org.apache.paimon.data.BinaryRow; + +import org.junit.jupiter.api.Test; + +import static org.assertj.core.api.Assertions.assertThat; + +/** Tests for {@link PartitionEntry#merge(PartitionEntry)}. */ +public class PartitionEntryTest { + + private static final BinaryRow PARTITION = BinaryRow.EMPTY_ROW; + + /** + * Creates a PartitionEntry with the given fileCount, totalBuckets, and creation time. + * recordCount and fileSizeInBytes are set to fileCount for simplicity. + */ + private static PartitionEntry entry(long fileCount, int totalBuckets, long creationTime) { + return new PartitionEntry( + PARTITION, fileCount, fileCount, fileCount, creationTime, totalBuckets); + } + + // ------------------------------------------------------------------------- + // Tests for totalBuckets selection based on lastFileCreationTime + // ------------------------------------------------------------------------- + + @Test + public void testMergeTakesTotalBucketsFromNewerEntry() { + // Old files (2 buckets, earlier creation time) merged with new files (4 buckets, later). + // totalBuckets should come from the newer entry. + PartitionEntry old = entry(3, 2, 1000L); + PartitionEntry newer = entry(3, 4, 2000L); + + PartitionEntry result = old.merge(newer); + assertThat(result.totalBuckets()).isEqualTo(4); + assertThat(result.lastFileCreationTime()).isEqualTo(2000L); + assertThat(result.fileCount()).isEqualTo(6); + } + + @Test + public void testMergeOrderDoesNotAffectTotalBuckets() { + // Regardless of whether old.merge(newer) or newer.merge(old) is called, + // the result must always take totalBuckets from the entry with the later creation time. + PartitionEntry old = entry(3, 2, 1000L); + PartitionEntry newer = entry(3, 4, 2000L); + + PartitionEntry result1 = old.merge(newer); + PartitionEntry result2 = newer.merge(old); + + assertThat(result1.totalBuckets()).isEqualTo(4); + assertThat(result2.totalBuckets()).isEqualTo(4); + assertThat(result1.lastFileCreationTime()).isEqualTo(2000L); + assertThat(result2.lastFileCreationTime()).isEqualTo(2000L); + } + + @Test + public void testMergeWithDeleteEntryPreservesNewerTotalBuckets() { + // Simulates the scenario after INSERT OVERWRITE with rescale: + // - original ADD entries (2 buckets, time=1000) still present in base manifest + // - DELETE entries for old files (2 buckets, time=1000) in delta manifest + // - new ADD entries (4 buckets, time=2000) in delta manifest + // + // The merged entry should have totalBuckets=4 (from the newest files). + PartitionEntry originalAdd = entry(3, 2, 1000L); // original ADD (base manifest) + PartitionEntry deleteOld = entry(-3, 2, 1000L); // DELETE old files (same timestamp) + PartitionEntry newAdd = entry(3, 4, 2000L); // new ADD after overwrite + + // Simulate concurrent processing in any order (all 6 permutations produce same result) + PartitionEntry r1 = originalAdd.merge(deleteOld).merge(newAdd); + PartitionEntry r2 = originalAdd.merge(newAdd).merge(deleteOld); + PartitionEntry r3 = deleteOld.merge(originalAdd).merge(newAdd); + PartitionEntry r4 = deleteOld.merge(newAdd).merge(originalAdd); + PartitionEntry r5 = newAdd.merge(originalAdd).merge(deleteOld); + PartitionEntry r6 = newAdd.merge(deleteOld).merge(originalAdd); + + for (PartitionEntry r : new PartitionEntry[] {r1, r2, r3, r4, r5, r6}) { + assertThat(r.totalBuckets()) + .as("totalBuckets should be 4 regardless of merge order") + .isEqualTo(4); + assertThat(r.fileCount()) + .as("net fileCount should be 3 (original 3 files remain visible)") + .isEqualTo(3); + assertThat(r.lastFileCreationTime()).isEqualTo(2000L); + } + } + + @Test + public void testMergeWithEqualCreationTimeTakesFirstTotalBuckets() { + // When creation times are equal, the receiver's (this) totalBuckets is used. + // This is the tie-breaking behavior: the entry already accumulated takes precedence. + PartitionEntry a = entry(1, 2, 1000L); + PartitionEntry b = entry(1, 4, 1000L); + + PartitionEntry result = a.merge(b); + assertThat(result.totalBuckets()) + .isEqualTo(2); // 'a' wins tie (lastCreationTime >= entry's) + assertThat(result.fileCount()).isEqualTo(2); + } + + @Test + public void testMergeAggregatesCountsCorrectly() { + PartitionEntry a = entry(5, 4, 1000L); + PartitionEntry b = entry(3, 4, 2000L); + + PartitionEntry result = a.merge(b); + assertThat(result.fileCount()).isEqualTo(8); + assertThat(result.recordCount()).isEqualTo(8); + assertThat(result.fileSizeInBytes()).isEqualTo(8); + assertThat(result.totalBuckets()).isEqualTo(4); + assertThat(result.lastFileCreationTime()).isEqualTo(2000L); + } +} diff --git a/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java b/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java index 1470f92a3747..16ae71f6cdaa 100644 --- a/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/operation/FileSystemWriteRestoreTest.java @@ -19,236 +19,219 @@ package org.apache.paimon.operation; import org.apache.paimon.CoreOptions; -import org.apache.paimon.catalog.Identifier; import org.apache.paimon.data.BinaryRow; import org.apache.paimon.data.BinaryRowWriter; import org.apache.paimon.data.GenericRow; +import org.apache.paimon.fs.Path; +import org.apache.paimon.fs.local.LocalFileIO; +import org.apache.paimon.options.Options; import org.apache.paimon.schema.Schema; +import org.apache.paimon.schema.SchemaManager; +import org.apache.paimon.schema.SchemaUtils; +import org.apache.paimon.schema.TableSchema; +import org.apache.paimon.table.CatalogEnvironment; import org.apache.paimon.table.FileStoreTable; -import org.apache.paimon.table.TableTestBase; -import org.apache.paimon.table.sink.CommitMessage; -import org.apache.paimon.table.sink.CommitMessageImpl; +import org.apache.paimon.table.FileStoreTableFactory; +import org.apache.paimon.table.sink.StreamTableCommit; import org.apache.paimon.table.sink.StreamTableWrite; -import org.apache.paimon.table.sink.StreamWriteBuilder; -import org.apache.paimon.table.sink.TableCommitImpl; +import org.apache.paimon.types.DataType; import org.apache.paimon.types.DataTypes; +import org.apache.paimon.types.RowType; import org.junit.jupiter.api.Test; +import org.junit.jupiter.api.io.TempDir; +import java.util.Arrays; import java.util.Collections; -import java.util.List; import java.util.UUID; import static org.assertj.core.api.Assertions.assertThat; /** - * Tests for {@link FileSystemWriteRestore} covering the scenario where a partition has a different - * bucket count than the table default (e.g., after a rescale operation). + * Tests for {@link FileSystemWriteRestore}, covering the {@code totalBuckets} resolution logic for + * both empty and non-empty buckets across partitioned and unpartitioned tables. + * + *

    When restoring files for a {@code (partition, bucket)} that has no existing data files, there + * are no manifest entries to derive {@code totalBuckets} from. For partitioned tables, {@link + * WriteRestore#extractTotalBuckets} falls back to {@link + * org.apache.paimon.table.sink.PartitionBucketMapping} to correctly return the per-partition bucket + * count (e.g. after a rescale). For unpartitioned tables, {@code null} is returned so the write + * path falls back to {@code numBuckets} and the committer-side mismatch check still fires. */ -public class FileSystemWriteRestoreTest extends TableTestBase { - - /** - * Scenario: - * - *

      - *
    • Table bucket (default): 32 - *
    • Partition A bucket: 2 (rescaled) - *
    • Partition A bucket=0: has data files, totalBuckets=2 - *
    • Partition A bucket=1: no data files - *
    - * - *

    When restoring bucket=1 of partition A (empty bucket), the returned {@code totalBuckets} - * must be 2 (from the per-partition mapping), not 32 (the table default). - */ +public class FileSystemWriteRestoreTest { + + @TempDir java.nio.file.Path tempDir; + + private static final RowType ROW_TYPE = + RowType.of( + new DataType[] {DataTypes.INT(), DataTypes.INT(), DataTypes.BIGINT()}, + new String[] {"pt", "k", "v"}); + @Test - public void testEmptyBucketUsesPartitionBucketCount() throws Exception { - // Table default: 32 buckets - int tableBuckets = 32; - // Partition A was rescaled to 2 buckets - int partitionBuckets = 2; - - Identifier identifier = new Identifier("db", "table"); - catalog.createDatabase("db", false); - Schema schema = - Schema.newBuilder() - .column("pt", DataTypes.INT()) - .column("k", DataTypes.INT()) - .column("v", DataTypes.INT()) - .primaryKey("pt", "k") - .partitionKeys("pt") - .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) - .build(); - catalog.createTable(identifier, schema, false); - FileStoreTable table = getTable(identifier); - - // Write data to partition A, bucket=0 with totalBuckets=2 (simulating a rescaled - // partition). We write normally first, then re-wrap the commit message to override - // totalBuckets. - BinaryRow partitionA = partitionRow(1); - String commitUser = UUID.randomUUID().toString(); - StreamWriteBuilder writeBuilder = table.newStreamWriteBuilder().withCommitUser(commitUser); - List messages; - try (StreamTableWrite write = writeBuilder.newWrite()) { - // Write to partition=1, key=1, value=100 with bucket=0 explicitly - write.write(GenericRow.of(1, 1, 100), 0); - messages = write.prepareCommit(false, 0); - } + public void testEmptyBucketUsesPartitionBucketMapping() throws Exception { + // Build a table with default bucket=4 and write data into partition 1. + // Some buckets within partition 1 will end up with files (bucket 0 OR + // bucket 1, depending on hash); the OTHER bucket will be empty. Then + // "rescale" the table-level default to 32 (without rewriting partition 1) + // and ask the WriteRestore for an empty bucket. It must return + // totalBuckets=4 (the partition's actual bucket count), NOT 32 (the new + // table default). + FileStoreTable table = createPartitionedPkTable(4); - // Rewrap the commit message so that totalBuckets=2 (the rescaled partition bucket count) - CommitMessageImpl original = (CommitMessageImpl) messages.get(0); - CommitMessageImpl rescaled = - new CommitMessageImpl( - original.partition(), - original.bucket(), - partitionBuckets, - original.newFilesIncrement(), - original.compactIncrement()); - - try (TableCommitImpl commit = table.newCommit(commitUser)) { - commit.commit(0, Collections.singletonList(rescaled)); - } + // Write enough rows to populate at least one bucket within partition 1. + commitOneRow(table, /* pt */ 1, /* k */ 1); + commitOneRow(table, /* pt */ 1, /* k */ 2); + + // Find an empty bucket in partition 1 by inspecting the existing files. + int emptyBucket = findEmptyBucket(table, 1, /* totalBuckets */ 4); + + // Simulate a rescale by raising the table-level default bucket count + // (without rewriting existing files). Existing manifest entries still + // carry totalBuckets=4. + table = withBucket(table, 32); - // Now create the FileSystemWriteRestore for this table - FileStoreTable freshTable = getTable(identifier); - FileSystemWriteRestore writeRestore = newWriteRestore(freshTable); - - // Restore bucket=0 (has data files): totalBuckets should be 2 - RestoreFiles restored0 = writeRestore.restoreFiles(partitionA, 0, false, false); - assertThat(restored0.totalBuckets()) - .as("bucket=0 (has files) should use partition bucket count, not table default") - .isEqualTo(partitionBuckets); - assertThat(restored0.dataFiles()).isNotEmpty(); - - // Restore bucket=1 (empty bucket): totalBuckets should ALSO be 2, not 32 - RestoreFiles restored1 = writeRestore.restoreFiles(partitionA, 1, false, false); - assertThat(restored1.totalBuckets()) - .as("bucket=1 (empty bucket) should use partition bucket count, not table default") - .isEqualTo(partitionBuckets); - assertThat(restored1.dataFiles()).isEmpty(); + WriteRestore restore = newWriteRestore(table); + + RestoreFiles restored = restore.restoreFiles(binaryRow(1), emptyBucket, false, false); + + assertThat(restored.totalBuckets()) + .as( + "Empty (partition 1, bucket %d): totalBuckets must be inferred from " + + "PartitionBucketMapping (4), not the new table default (32).", + emptyBucket) + .isEqualTo(4); + assertThat(restored.dataFiles()).isNullOrEmpty(); } - /** - * Sanity check: a partition that has never been rescaled uses the table default bucket count. - */ @Test - public void testPartitionWithDefaultBucketCount() throws Exception { - int tableBuckets = 32; - - Identifier identifier = new Identifier("db2", "table"); - catalog.createDatabase("db2", false); - Schema schema = - Schema.newBuilder() - .column("pt", DataTypes.INT()) - .column("k", DataTypes.INT()) - .column("v", DataTypes.INT()) - .primaryKey("pt", "k") - .partitionKeys("pt") - .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) - .build(); - catalog.createTable(identifier, schema, false); - FileStoreTable table = getTable(identifier); - - // Write data using the default bucket count (no totalBuckets override) - write(table, GenericRow.of(1, 1, 100)); - - FileStoreTable freshTable = getTable(identifier); - FileSystemWriteRestore writeRestore = newWriteRestore(freshTable); - - BinaryRow partitionA = partitionRow(1); - - // Restore bucket=0 (has data with default totalBuckets=32) - RestoreFiles restored = writeRestore.restoreFiles(partitionA, 0, false, false); - assertThat(restored.totalBuckets()) - .as("partition with default bucket count should return table bucket count") - .isEqualTo(tableBuckets); + public void testEmptyBucketInUnseenPartitionUsesDefault() throws Exception { + // For an entirely unseen partition (no files anywhere), no per-partition + // mapping exists and PartitionBucketMapping.resolveNumBuckets falls back to + // the table's default bucket count. + FileStoreTable table = createPartitionedPkTable(8); + commitOneRow(table, 1, 100); // ensures the snapshot exists + + WriteRestore restore = newWriteRestore(table); + RestoreFiles restored = restore.restoreFiles(binaryRow(/* unseen */ 999), 0, false, false); + + assertThat(restored.totalBuckets()).isEqualTo(8); + assertThat(restored.dataFiles()).isNullOrEmpty(); } - /** - * Scenario with two partitions: partition A rescaled to 2 buckets, partition B uses default 32. - * Each partition's empty buckets must return their own bucket count. - */ @Test - public void testMixedPartitionsWithDifferentBucketCounts() throws Exception { - int tableBuckets = 32; - int partitionABuckets = 2; - - Identifier identifier = new Identifier("db3", "table"); - catalog.createDatabase("db3", false); - Schema schema = - Schema.newBuilder() - .column("pt", DataTypes.INT()) - .column("k", DataTypes.INT()) - .column("v", DataTypes.INT()) - .primaryKey("pt", "k") - .partitionKeys("pt") - .option(CoreOptions.BUCKET.key(), String.valueOf(tableBuckets)) - .build(); - catalog.createTable(identifier, schema, false); - FileStoreTable table = getTable(identifier); - - String commitUser = UUID.randomUUID().toString(); - - // Write partition A, bucket=0 with rescaled totalBuckets=2 - StreamWriteBuilder writeBuilder = table.newStreamWriteBuilder().withCommitUser(commitUser); - List messagesA; - try (StreamTableWrite write = writeBuilder.newWrite()) { - write.write(GenericRow.of(1, 1, 100), 0); - messagesA = write.prepareCommit(false, 0); - } - CommitMessageImpl originalA = (CommitMessageImpl) messagesA.get(0); - CommitMessageImpl rescaledA = - new CommitMessageImpl( - originalA.partition(), - originalA.bucket(), - partitionABuckets, - originalA.newFilesIncrement(), - originalA.compactIncrement()); - try (TableCommitImpl commit = table.newCommit(commitUser)) { - commit.commit(0, Collections.singletonList(rescaledA)); - } + public void testNonEmptyBucketReportsManifestTotalBuckets() throws Exception { + // Sanity test: when a bucket has files, totalBuckets must come from the + // manifest entries (not from the fallback path). This guards against + // accidentally always overriding totalBuckets via PartitionBucketMapping. + FileStoreTable table = createPartitionedPkTable(2); + commitOneRow(table, 1, 1); + commitOneRow(table, 1, 2); + + // Locate a non-empty bucket within partition 1. + int nonEmptyBucket = findNonEmptyBucket(table, 1, 2); + + // Change the table default to ensure the returned totalBuckets is from the + // manifest entry, not the schema. + table = withBucket(table, 32); + + WriteRestore restore = newWriteRestore(table); + RestoreFiles restored = restore.restoreFiles(binaryRow(1), nonEmptyBucket, false, false); + + assertThat(restored.totalBuckets()).isEqualTo(2); + assertThat(restored.dataFiles()).isNotEmpty(); + } - // Write partition B, bucket=0 with default totalBuckets=32 - try (StreamTableWrite write = writeBuilder.newWrite()) { - write.write(GenericRow.of(2, 1, 200), 0); - List messagesB = write.prepareCommit(false, 1); - try (TableCommitImpl commit = table.newCommit(commitUser)) { - commit.commit(1, messagesB); - } - } + // ------------------------------------------------------------------------ + // helpers + // ------------------------------------------------------------------------ + + private FileStoreTable createPartitionedPkTable(int bucket) throws Exception { + Path path = new Path(tempDir.toString()); + Options options = new Options(); + options.set(CoreOptions.PATH, path.toString()); + options.set(CoreOptions.BUCKET, bucket); + + TableSchema tableSchema = + SchemaUtils.forceCommit( + new SchemaManager(LocalFileIO.create(), path), + new Schema( + ROW_TYPE.getFields(), + Collections.singletonList("pt"), + Arrays.asList("pt", "k"), + options.toMap(), + "")); + + return FileStoreTableFactory.create( + LocalFileIO.create(), path, tableSchema, CatalogEnvironment.empty()); + } - FileStoreTable freshTable = getTable(identifier); - FileSystemWriteRestore writeRestore = newWriteRestore(freshTable); - - BinaryRow partitionA = partitionRow(1); - BinaryRow partitionB = partitionRow(2); - - // Partition A: bucket=1 (empty) should use 2, not 32 - RestoreFiles restoredA1 = writeRestore.restoreFiles(partitionA, 1, false, false); - assertThat(restoredA1.totalBuckets()) - .as("partition A empty bucket should use rescaled partition bucket count 2") - .isEqualTo(partitionABuckets); - assertThat(restoredA1.dataFiles()).isEmpty(); - - // Partition B: bucket=0 (has files) should use 32 - RestoreFiles restoredB0 = writeRestore.restoreFiles(partitionB, 0, false, false); - assertThat(restoredB0.totalBuckets()) - .as("partition B should use table default bucket count 32") - .isEqualTo(tableBuckets); - assertThat(restoredB0.dataFiles()).isNotEmpty(); + private FileStoreTable withBucket(FileStoreTable table, int newBucket) { + Options options = new Options(table.options()); + options.set(CoreOptions.BUCKET, newBucket); + return table.copy(table.schema().copy(options.toMap())); } - private static FileSystemWriteRestore newWriteRestore(FileStoreTable table) { + private WriteRestore newWriteRestore(FileStoreTable table) { return new FileSystemWriteRestore( table.store().options(), - table.store().snapshotManager(), + table.snapshotManager(), table.store().newScan(), table.store().newIndexFileHandler()); } - private BinaryRow partitionRow(int partitionValue) { + private void commitOneRow(FileStoreTable table, int pt, int k) throws Exception { + String user = UUID.randomUUID().toString(); + Long latest = table.snapshotManager().latestSnapshotId(); + long id = latest == null ? 0L : latest; + try (StreamTableWrite write = table.newWrite(user); + StreamTableCommit commit = table.newCommit(user)) { + write.write(GenericRow.of(pt, k, (long) k)); + commit.commit(id, write.prepareCommit(true, id)); + } + } + + /** Returns a bucket id (0..totalBuckets-1) that has no data files within the partition. */ + private int findEmptyBucket(FileStoreTable table, int pt, int totalBuckets) throws Exception { + BinaryRow partition = binaryRow(pt); + for (int b = 0; b < totalBuckets; b++) { + int bucket = b; + boolean nonEmpty = + table.newSnapshotReader() + .withPartitionFilter(Collections.singletonList(partition)) + .withBucket(bucket).read().dataSplits().stream() + .anyMatch(s -> !s.dataFiles().isEmpty()); + if (!nonEmpty) { + return bucket; + } + } + throw new IllegalStateException( + "Could not find an empty bucket in partition " + + pt + + " (every bucket has files); test scenario could not be set up."); + } + + /** Returns a bucket id (0..totalBuckets-1) that has at least one data file. */ + private int findNonEmptyBucket(FileStoreTable table, int pt, int totalBuckets) + throws Exception { + BinaryRow partition = binaryRow(pt); + for (int b = 0; b < totalBuckets; b++) { + int bucket = b; + boolean nonEmpty = + table.newSnapshotReader() + .withPartitionFilter(Collections.singletonList(partition)) + .withBucket(bucket).read().dataSplits().stream() + .anyMatch(s -> !s.dataFiles().isEmpty()); + if (nonEmpty) { + return bucket; + } + } + throw new IllegalStateException("Could not find a non-empty bucket in partition " + pt); + } + + private static BinaryRow binaryRow(int pt) { BinaryRow row = new BinaryRow(1); BinaryRowWriter writer = new BinaryRowWriter(row); - writer.writeInt(0, partitionValue); + writer.writeInt(0, pt); writer.complete(); return row; } diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java index 6d3dea1eddef..b8c49789d635 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java @@ -152,12 +152,8 @@ public synchronized ScanCoordinationResponse scan(ScanCoordinationRequest reques List entries = scan.withPartitionBucket(partition, bucket).plan().files(); List restoreFiles = WriteRestore.extractDataFiles(entries); - - // Resolve the totalBuckets from the partition-level mapping rather than - // using the bucket entries, because a partition may have been rescaled to a - // different totalBuckets. If we used the bucket entries for an empty bucket, - // we would incorrectly return the table default totalBuckets for that partition. - Integer totalBuckets = partitionBucketMapping.resolveNumBuckets(partition); + Integer totalBuckets = + WriteRestore.extractTotalBuckets(entries, partition, partitionBucketMapping); IndexFileMeta dynamicBucketIndex = null; if (request.scanDynamicBucketIndex()) { From 9c784bf0640fbaf3ca8b674b01aabdf200ede4ff Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Tue, 26 May 2026 09:37:45 +1000 Subject: [PATCH 6/7] Fix merge commutativity, fail fast on scan error, and add streaming restart note for per-partition bucket rescaling --- docs/docs/maintenance/rescale-bucket.md | 6 ++++ .../paimon/manifest/PartitionEntry.java | 16 +++++++--- .../table/SchemaBucketFileStoreTable.java | 8 +++++ .../table/sink/PartitionBucketMapping.java | 31 +++++++++---------- .../paimon/manifest/PartitionEntryTest.java | 19 +++++++----- .../sink/PartitionBucketMappingTest.java | 16 ++++++++++ .../coordinator/TableWriteCoordinator.java | 5 +++ 7 files changed, 72 insertions(+), 29 deletions(-) diff --git a/docs/docs/maintenance/rescale-bucket.md b/docs/docs/maintenance/rescale-bucket.md index 74889941f7ff..a814f28a60a2 100644 --- a/docs/docs/maintenance/rescale-bucket.md +++ b/docs/docs/maintenance/rescale-bucket.md @@ -68,6 +68,12 @@ Please note that After these operations, partition `dt=2022-01-01` uses 4 buckets, `dt=2022-01-02` uses 8 buckets, and any new partitions will use the latest table-level default (8 buckets in this case). - During overwrite period, make sure there are no other jobs writing the same table/partition. +- **Streaming jobs must be restarted after rescaling a partition.** The per-partition bucket mapping + is loaded once when the streaming job starts (from the manifest files at that point in time). If a + partition is rescaled while the streaming job is running, the job will continue routing rows using + the old bucket count for that partition, which can cause rows to land in wrong buckets and lead to + data correctness issues. The recommended workflow is: suspend the streaming job with a savepoint → + perform the rescale overwrite → restart from the savepoint. ## Use Case diff --git a/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java b/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java index 43ba6d1a0f48..b34fa528f05b 100644 --- a/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java +++ b/paimon-core/src/main/java/org/apache/paimon/manifest/PartitionEntry.java @@ -83,20 +83,26 @@ public int totalBuckets() { } public PartitionEntry merge(PartitionEntry entry) { + PartitionEntry newer = entry.lastFileCreationTime >= lastFileCreationTime ? entry : this; + PartitionEntry older = newer == entry ? this : entry; + // Use the totalBuckets from the most recently created file. This correctly handles // the case where a partition has been overwritten with a different bucket count: the // newer files carry the new totalBuckets, and their creation time is always later. - long newLastCreationTime = Math.max(lastFileCreationTime, entry.lastFileCreationTime); + // When timestamps are equal (e.g., two files written in the same millisecond with + // different bucket counts), we take the larger totalBuckets value. This makes merge + // commutative and associative — a.merge(b) == b.merge(a) int newTotalBuckets = - lastFileCreationTime >= entry.lastFileCreationTime - ? totalBuckets - : entry.totalBuckets; + newer.lastFileCreationTime == older.lastFileCreationTime + ? Math.max(newer.totalBuckets, older.totalBuckets) + : newer.totalBuckets; + return new PartitionEntry( partition, recordCount + entry.recordCount, fileSizeInBytes + entry.fileSizeInBytes, fileCount + entry.fileCount, - newLastCreationTime, + newer.lastFileCreationTime, newTotalBuckets); } diff --git a/paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java b/paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java index 88e2dfbb8894..e730c8ba505e 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/SchemaBucketFileStoreTable.java @@ -65,6 +65,14 @@ public TableWriteImpl newWrite(String commitUser, @Nullable Integer writeId) return wrapped().newWrite(commitUser, writeId, createRowKeyExtractor()); } + @Override + public TableWriteImpl newWrite( + String commitUser, @Nullable Integer writeId, RowKeyExtractor rowKeyExtractor) { + // Always use the schema-bucket-based extractor; ignore the caller-supplied extractor + // to ensure consistent per-partition bucket routing even when called via the 3-arg form. + return wrapped().newWrite(commitUser, writeId, createRowKeyExtractor()); + } + @Override public FileStoreTable copy(Map dynamicOptions) { return new SchemaBucketFileStoreTable(wrapped().copy(dynamicOptions)); diff --git a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java index a1d98c24af7e..5e39c29a998c 100644 --- a/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java +++ b/paimon-core/src/main/java/org/apache/paimon/table/sink/PartitionBucketMapping.java @@ -83,8 +83,7 @@ public PartitionBucketMapping( *

    For partitioned tables, the method reads {@link * org.apache.paimon.manifest.PartitionEntry}s, which aggregate manifest entries per partition * during the scan and therefore have a much smaller memory footprint than loading all data file - * entries. If the scan fails for any reason, a fallback mapping with only the default bucket - * count is returned. + * entries. Any scan failure is propagated to the caller. * * @param table the {@link FileStoreTable} to load the mapping from * @return a {@link PartitionBucketMapping} reflecting the current bucket layout of the table @@ -98,23 +97,21 @@ public static PartitionBucketMapping loadFromTable(FileStoreTable table) { } public static PartitionBucketMapping loadFromScan(FileStoreScan scan, int defaultBuckets) { - try { - List partitionEntries = scan.readPartitionEntries(); - Map partitionBucketMap = new HashMap<>(); - for (PartitionEntry entry : partitionEntries) { - int totalBuckets = entry.totalBuckets(); - // Only store partitions whose bucket count differs from the default. - // This keeps the map empty for partitions that have never been rescaled, - // avoiding per-partition BinaryRow copies and Integer allocations entirely. - if (totalBuckets > 0 && totalBuckets != defaultBuckets) { - partitionBucketMap.put(entry.partition().copy(), totalBuckets); - } - } - - return new PartitionBucketMapping(defaultBuckets, partitionBucketMap); - } catch (Exception e) { + if (scan == null) { return new PartitionBucketMapping(defaultBuckets, Collections.emptyMap()); } + List partitionEntries = scan.readPartitionEntries(); + Map partitionBucketMap = new HashMap<>(); + for (PartitionEntry entry : partitionEntries) { + int totalBuckets = entry.totalBuckets(); + // Only store partitions whose bucket count differs from the default. + // This keeps the map empty for partitions that have never been rescaled, + // avoiding per-partition BinaryRow copies and Integer allocations entirely. + if (totalBuckets > 0 && totalBuckets != defaultBuckets) { + partitionBucketMap.put(entry.partition().copy(), totalBuckets); + } + } + return new PartitionBucketMapping(defaultBuckets, partitionBucketMap); } /** diff --git a/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java b/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java index 526221690a2b..8cc07f65d024 100644 --- a/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/manifest/PartitionEntryTest.java @@ -103,16 +103,21 @@ public void testMergeWithDeleteEntryPreservesNewerTotalBuckets() { } @Test - public void testMergeWithEqualCreationTimeTakesFirstTotalBuckets() { - // When creation times are equal, the receiver's (this) totalBuckets is used. - // This is the tie-breaking behavior: the entry already accumulated takes precedence. + public void testMergeWithEqualCreationTimeIsCommutative() { + // When creation times are equal, merge must be commutative: a.merge(b) == b.merge(a). + // The tie-break takes the larger totalBuckets so that the parallel, non-deterministic + // aggregation in readPartitionEntries() always produces the same result regardless of + // manifest processing order. PartitionEntry a = entry(1, 2, 1000L); PartitionEntry b = entry(1, 4, 1000L); - PartitionEntry result = a.merge(b); - assertThat(result.totalBuckets()) - .isEqualTo(2); // 'a' wins tie (lastCreationTime >= entry's) - assertThat(result.fileCount()).isEqualTo(2); + PartitionEntry ab = a.merge(b); + PartitionEntry ba = b.merge(a); + + assertThat(ab.totalBuckets()).isEqualTo(4); // max(2, 4) = 4 + assertThat(ba.totalBuckets()).isEqualTo(4); // max(4, 2) = 4, commutative + assertThat(ab.fileCount()).isEqualTo(2); + assertThat(ba.fileCount()).isEqualTo(2); } @Test diff --git a/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java b/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java index f6958720c30f..9fd85e348ecc 100644 --- a/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/table/sink/PartitionBucketMappingTest.java @@ -19,13 +19,16 @@ package org.apache.paimon.table.sink; import org.apache.paimon.data.BinaryRow; +import org.apache.paimon.operation.FileStoreScan; import org.junit.jupiter.api.Test; +import org.mockito.Mockito; import java.util.HashMap; import java.util.Map; import static org.assertj.core.api.Assertions.assertThat; +import static org.assertj.core.api.Assertions.assertThatThrownBy; /** Tests for {@link PartitionBucketMapping}. */ public class PartitionBucketMappingTest { @@ -60,6 +63,19 @@ public void testExplicitPartitionMapping() { assertThat(mapping.resolveNumBuckets(partC)).isEqualTo(16); } + @Test + public void testLoadFromScanPropagatesException() { + // Simulate a scan that throws (e.g. corrupted manifest, transient I/O error). + // loadFromScan must fail fast so the job does not silently write to wrong buckets. + FileStoreScan failingScan = Mockito.mock(FileStoreScan.class); + Mockito.when(failingScan.readPartitionEntries()) + .thenThrow(new RuntimeException("simulated manifest scan failure")); + + assertThatThrownBy(() -> PartitionBucketMapping.loadFromScan(failingScan, 8)) + .isInstanceOf(RuntimeException.class) + .hasMessageContaining("simulated manifest scan failure"); + } + private static BinaryRow partition(int value) { return BinaryRow.singleColumn(value); } diff --git a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java index b8c49789d635..a31079f1aa63 100644 --- a/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java +++ b/paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/TableWriteCoordinator.java @@ -199,6 +199,11 @@ public void checkpoint() { private void loadPartitionBucketMapping() { int defaultNumBuckets = table.schema().numBuckets(); + // Note: `scan` is shared between this method (called during refresh/checkpoint) and the + // `scan(ScanCoordinationRequest)` method (which calls scan.withPartitionBucket(...)). + // Both callers always invoke scan.withSnapshot(...) before using the scan, so the shared + // state is safe. The partition-bucket filter set by withPartitionBucket is applied per + // plan() call and does not bleed across invocations. this.partitionBucketMapping = PartitionBucketMapping.loadFromScan(scan, defaultNumBuckets); } From 2360910c5497441ebfd5ef3a9722466da873ef62 Mon Sep 17 00:00:00 2001 From: Mike Dias Date: Tue, 26 May 2026 15:42:01 +1000 Subject: [PATCH 7/7] Improve how we test the BUCKET_APPEND_ORDERED behaviour --- .../table/AppendOnlySimpleTableTest.java | 61 ++++++++++++------- 1 file changed, 39 insertions(+), 22 deletions(-) diff --git a/paimon-core/src/test/java/org/apache/paimon/table/AppendOnlySimpleTableTest.java b/paimon-core/src/test/java/org/apache/paimon/table/AppendOnlySimpleTableTest.java index c91dbb4c8e4f..cbd5fee57f87 100644 --- a/paimon-core/src/test/java/org/apache/paimon/table/AppendOnlySimpleTableTest.java +++ b/paimon-core/src/test/java/org/apache/paimon/table/AppendOnlySimpleTableTest.java @@ -99,6 +99,7 @@ import java.util.Optional; import java.util.PriorityQueue; import java.util.Random; +import java.util.Set; import java.util.UUID; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors; @@ -155,16 +156,16 @@ public void testOverwriteSameFiles() throws Exception { } @Test - public void testBucketedAppendTableWriteWithInit() throws Exception { - innerTestBucketedAppendTableWriteInit(true); + public void testBucketedAppendOrderedSequenceNumbers() throws Exception { + innerTestBucketedAppendSequenceNumbers(true); } @Test - public void testBucketedAppendTableWriteNoInit() throws Exception { - innerTestBucketedAppendTableWriteInit(false); + public void testBucketedAppendUnorderedSequenceNumbers() throws Exception { + innerTestBucketedAppendSequenceNumbers(false); } - public void innerTestBucketedAppendTableWriteInit(boolean ordered) throws Exception { + public void innerTestBucketedAppendSequenceNumbers(boolean ordered) throws Exception { FileStoreTable table = createFileStoreTable( options -> { @@ -176,31 +177,47 @@ public void innerTestBucketedAppendTableWriteInit(boolean ordered) throws Except BatchWriteBuilder writeBuilder = table.newBatchWriteBuilder(); - // 1. first write + // 1. first write - use a=1 so both batches land in the same bucket try (BatchTableWrite write = writeBuilder.newWrite(); BatchTableCommit commit = writeBuilder.newCommit()) { write.write(rowData(1, 10, 100L)); commit.commit(write.prepareCommit()); } - // 2. delete all manifests - ManifestList manifestList = table.store().manifestListFactory().create(); - ManifestFile manifestFile = table.store().manifestFileFactory().create(); - List manifests = - manifestList.readAllManifests(table.latestSnapshot().get()); - for (ManifestFileMeta manifest : manifests) { - manifestFile.delete(manifest.fileName()); + // collect sequence numbers from batch 1 + List batch1Files = + table.newReadBuilder().newScan().plan().splits().stream() + .flatMap(s -> ((DataSplit) s).dataFiles().stream()) + .collect(Collectors.toList()); + long batch1MaxSeq = + batch1Files.stream().mapToLong(DataFileMeta::maxSequenceNumber).max().getAsLong(); + Set batch1FileNames = + batch1Files.stream().map(DataFileMeta::fileName).collect(Collectors.toSet()); + + // 2. second write - same a=1 value ensures same bucket as batch 1 + try (BatchTableWrite write = writeBuilder.newWrite(); + BatchTableCommit commit = writeBuilder.newCommit()) { + write.write(rowData(1, 20, 200L)); + commit.commit(write.prepareCommit()); } - // 3. check new write - try (BatchTableWrite write = writeBuilder.newWrite()) { - if (ordered) { - assertThatThrownBy(() -> write.write(rowData(1, 10, 100L))) - .hasMessageContaining("FileNotFoundException"); - } else { - // no exception - write.write(rowData(1, 10, 100L)); - } + // collect sequence numbers from batch 2 only (exclude batch 1 files by name) + List batch2Files = + table.newReadBuilder().newScan().plan().splits().stream() + .flatMap(s -> ((DataSplit) s).dataFiles().stream()) + .filter(f -> !batch1FileNames.contains(f.fileName())) + .collect(Collectors.toList()); + long batch2MinSeq = + batch2Files.stream().mapToLong(DataFileMeta::minSequenceNumber).min().getAsLong(); + + if (ordered) { + // ordered mode always restores previous files and continues sequence numbers, + // so batch 2 sequence numbers are strictly greater than batch 1's + assertThat(batch2MinSeq).isGreaterThan(batch1MaxSeq); + } else { + // unordered+writeOnly mode skips restoring previous files (ignorePreviousFiles=true), + // so sequence numbers reset to 0 each session + assertThat(batch2MinSeq).isEqualTo(0L); } }