Glue Hudi get the freshly added or updated records

Question

Hello,

I'm using Hudi connector in Glue, first, I bulk inserted the initial dataset to Hudi table, I'm adding a daily incremental records and I can query them using Athena, what I'm trying to do is to get the newly added, updated or deleted records in a separate parquet file.

I've tried different approaches and configurations using both copy on write and merge on read tables but still can get the updates in a separate file.

I used these configurations in different combinations:

'className' : 'org.apache.hudi',
    'hoodie.datasource.hive_sync.use_jdbc': 'false',
    'hoodie.datasource.write.precombine.field': 'ts',
    'hoodie.datasource.write.recordkey.field': 'uuid',
    'hoodie.payload.event.time.field': 'ts',
    'hoodie.table.name': 'table_name',
    'hoodie.datasource.hive_sync.database': 'hudi_db',
    'hoodie.datasource.hive_sync.table': 'table_name',
    'hoodie.datasource.hive_sync.enable': 'false',
    # 'hoodie.datasource.write.partitionpath.field': 'date:SIMPLE',
    'hoodie.datasource.write.hive_style_partitioning': 'true',
    'hoodie.meta.sync.client.tool.class': 'org.apache.hudi.aws.sync.AwsGlueCatalogSyncTool',
    'hoodie.datasource.write.table.type': 'COPY_ON_WRITE',
    'path': 's3://path/to/output/',
    # 'hoodie.datasource.write.operation': 'bulk_insert',
    'hoodie.datasource.write.operation': 'upsert',
    # 'hoodie.datasource.hive_sync.partition_extractor_class': 'org.apache.hudi.hive.NonPartitionedExtractor',
    # 'hoodie.datasource.hive_sync.partition_extractor_class': 'org.apache.hudi.hive.MultiPartKeysValueExtractor',
    'hoodie.datasource.write.keygenerator.class': 'org.apache.hudi.keygen.NonpartitionedKeyGenerator',
    # 'hoodie.compaction.payload.class': 'org.apache.hudi.common.model.OverwriteWithLatestAvroPayload',
    # 'hoodie.cleaner.policy': 'KEEP_LATEST_COMMITS',
    'hoodie.cleaner.delete.bootstrap.base.file': 'true',
    "hoodie.index.type": "GLOBAL_BLOOM",
    'hoodie.file.index.enable': 'true',
    'hoodie.bloom.index.update.partition.path': 'true',
    'hoodie.bulkinsert.shuffle.parallelism': 1,
    # 'hoodie.datasource.write.keygenerator.class': 'org.apache.hudi.keygen.CustomKeyGenerator'

Thank you.

Answer

At this moment, Hudi does not have a feature to fit the use case you described. When use Copy on Write table, incremental updates will be compacted into base file in real time automatically. While with Merge on Read table, incremental updates are logged to delta files (in Avro format) and later compacted to base Parquet files. Reference: https://hudi.apache.org/docs/concepts/#table-types

A possible workaround, you can try implement your own data write and compact logic using custom ETL jobs on native Glue Catalog table instead of using hudi table.

I hope this helps!

A possible workaround, you can try implement your own data write and compact logic using custom ETL jobs on native Glue Catalog table instead of using hudi table.

I hope this helps!

Glue Hudi get the freshly added or updated records

Conteúdo relevante