Proposers

• Raymond Xu

Approvers

• Vinoth Chandar  :  [APPROVED/REQUESTED_INFO/REJECTED]
• Balaji Varadarajan  : APPROVED
• Nishith Agarwal :  [APPROVED/REQUESTED_INFO/REJECTED]

Status

Current state: Under Discussion 

Discussion thread: - here

JIRA:

Released: <Hudi Version>0.6.0

Abstract

A feature to snapshot a Hudi dataset and export the latest records to a set of external files (e.g., plain parquet files).

Background

The existing org.apache.hudi.utilities.HoodieSnapshotCopier performs a Hudi-to-Hudi copy that serves for backup purpose. To broaden the usability, the Copier could be potentially extended to perform exporting features to data formats, like plain parquet files, other than Hudi dataset.

Implementation

The proposed class is org.apache.hudi.utilities.HoodieSnapshotterHoodieSnapshotExporter , which serves as the main entry for snapshotting related work.

Definition of "Snapshot"

To snapshot is to get the most up-to-date records from a Hudi dataset at the query a particular point in time. Note that this could take longer for MOR tables as it involves merging the latest the data exported from MOR tables may not be the most up-to-date as RO query is used for retrieval, which omits the latest data in the log files.

Arguments

String partitionField = // from the argument
df.repartition(df.col(partitionField))
  .write()
  .partitionBy(partitionField)
  .parquet(outputPath);

Steps

1. Read
   • Output format "PARQUET": Leverage Regardless of output format, always leverage on org.apache.hudi.common.table.view.HoodieTableFileSystemView logic to get the latest records (RT query) to perform RO query for read
   • Specifically, data to be read is from the latest version of columnar files in the source dataset, up to the latest commit timeOutput format "HUDI_COPY": we don't need RT query. Instead, we just use RO query to copy the latest parquet files, like what the existing HoodieSnapshotCopier does
2. Transform
   • Output format "PARQUETparquet"
     • Stripe Hudi metadata
     • Allow user to provide a field to do simple Spark repartitioning
     • Allow user to provide a class to do custom repartitioning
   • No transformation is needed for output format "HUDI_COPYhudi"; just copy the original files, like what the existing HoodieSnapshotCopier does
3. Write
   • Just need to provide the output directory and Spark shall handle the rest.

Rollout/Adoption Plan

• No impact to existing users as this is a new independent utility tool.
• Once this feature is GA'ed, we can mark HoodieSnapshotCopier as deprecated and suggest user to switch to this tool, which provides equivalent copying features.

Test Plan

• Write similar tests like HoodieSnapshotCopier 
• When testing end-to-end, we are to verify
  • number of records are matched
  • later snapshot reflect the latest info from the original dataset

...