apache / apache/paimon

[Bug] Data loss occurs when Flink read paimon table with 'changelog-producer = full-compact' in 'latest-full' mode

Open
#2,787 0 comments 0 reactions 0 assignees View on GitHub
bug
Dominant language
Java
Stars
3.4k
Forks
1.4k
Avg merge
1d 11h
Merged PRs (30d)
396

Description

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/incubator-paimon/issues) and found nothing similar.

### Paimon version

0.7

### Compute Engine

Flink

### Minimal reproduce step

1. create paimon table with engine as "Deduplicate" and set "changelog-producer = none"
2. insert offline data to paimon table and waiting for the end
3. alter paimon table set "changelog-producer = full-compact"
4. writing stream data into paimon table
5. Start the flink job to read all paimon data, and found the amount of data is much less than actual data.

From the log of reading paimon, i found some data file haven't been read.

### What doesn't meet your expectations?

When i set "changelog-producer = full-compact" for paimon, data loss occurs when full data is read by flink job.

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Contributor guide

No contributing guide indexed for this repository

Research direction

Reproduce the issue using the listed Paimon 0.7 and Flink steps, including changing changelog-producer from none to full-compact and reading in latest-full mode. Inspect the read log for skipped data files and compare the Flink result with the actual table contents. Done means a full read returns all data without loss.

Written by the indexing model from the issue text.

Assessment

Tech stack
java
Domain
data-engineering, distributed-systems
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.