You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I searched in the issues and found nothing similar.
Motivation
Why we need this.
Currently , compact action is fullCompaction in batch mode, that will merge all base file with delta file and generates a new base file. After that, we will have two copies of the full data in storage (base_file1 + delta_file1 + base_file2).
But : Sometimes we just need to merge incremental data, we allow some reduction in read performance in exchange for storage space.
Solution
This will be implemented through 3 PRs :
step 1 : Refactor compact action to support extended compact type.
step 2:Compact action supports using --compact_strategy full/minor to decide which compaction will be triggered FullCompaction or UniversalCompaction.
step 3:Add a new Procedure universal_compact for spark and flink
Anything else?
No response
Are you willing to submit a PR?
I'm willing to submit a PR!
The text was updated successfully, but these errors were encountered:
LinMingQiang
changed the title
[Feature] Support universal compact action for batch mode.
[Feature] Support minor compact for dedicated compaction.
Nov 21, 2024
LinMingQiang
changed the title
[Feature] Support minor compact for dedicated compaction.
[Feature] Support minor compact strategy for dedicated compaction.
Nov 26, 2024
Search before asking
Motivation
Why we need this.
Currently , compact action is fullCompaction in batch mode, that will merge all base file with delta file and generates a new base file. After that, we will have two copies of the full data in storage (base_file1 + delta_file1 + base_file2).
But : Sometimes we just need to merge incremental data, we allow some reduction in read performance in exchange for storage space.
Solution
This will be implemented through 3 PRs :
step 1 : Refactor compact action to support extended compact type.
step 2:Compact action supports using
--compact_strategy full/minor
to decide which compaction will be triggered FullCompaction or UniversalCompaction.step 3:Add a new Procedure universal_compact for spark and flink
Anything else?
No response
Are you willing to submit a PR?
The text was updated successfully, but these errors were encountered: