[core] Fix checkpoint recovery failure for compacted changelog files #6173
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
JingsongLi
reviewed
Aug 29, 2025
| * Utility class for resolving compacted changelog file paths. | ||
| * | ||
| * <p>This class provides functionality to resolve fake compacted changelog file paths to their real | ||
| * file paths, following the same protocol as CompactedChangelogFormatReaderFactory. |
Contributor
There was a problem hiding this comment.
Can you also move logic codes into this class from CompactedChangelogFormatReaderFactory? It is better to keep protocol in one single class.
JingsongLi
reviewed
Aug 29, 2025
| */ | ||
| public static Path resolveCompactedChangelogPath(Path path) { | ||
| // Check if this is a compacted changelog file by its .cc- extension | ||
| String[] nameAndFormat = path.getName().split("\\."); |
Contributor
There was a problem hiding this comment.
Maybe just create a method isCompactedChangelogPath. And just check it is starting with compacted-changelog-?
Contributor
|
Thanks @uicosp for the contribution! Left comments. |
added 3 commits
September 5, 2025 12:15
Contributor
Author
|
Refactor:
|
added 4 commits
September 5, 2025 15:25
tsreaper
approved these changes
Sep 10, 2025
jerry-024
added a commit
to jerry-024/paimon
that referenced
this pull request
Sep 15, 2025
* upstream/master: (23 commits) [flink] Use Paimon format table read for flink (apache#6246) [core] Ensure system tables use the correct identifier for loadTableToken in RESTTokenFileIO. (apache#6247) [spark] Enhance v1 write merge schema test coverage (apache#6249) [spark] Eliminate duplicate convertLiteral invocations (apache#6250) [python] Fix failing to read 1000cols (apache#6244) [python] Expose CatalogFactory and Schema directly (apache#6243) [doc] Modify Python API to JVM free (apache#6242) [python] Fix multiple write brefore once commit (apache#6241) [core] Support push down branchesTable by branchName (apache#6231) [cdc] Fix PostgreSQL DECIMAL type conversion issue (apache#6239) [arrow] Optimize Arrow string write performance (apache#6240) [core] Fix checkpoint recovery failure for compacted changelog files (apache#6173) [core] RESTCatalog: add DLF OSS endpoint support and improve configuration merge (apache#6232) [core] fix RESTCatalog#listViews for system database (apache#6233) [core] Introduce 'ignore-update-before' to ignore UD only (apache#6235) [python] Fix DLF partition statistical error (apache#6237) [python] Add _VALUE_STATS_COLS param to fix parse wrong bytes (apache#6234) [ci] Rename to Python Check Code Style and Test [python] Rename binary row to generic row [hotfix] Remove methods in SchemaManager for SchemasTable ...
zhuyufeng0809
pushed a commit
to zhuyufeng0809/flink-table-store
that referenced
this pull request
Sep 18, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
[core] Fix checkpoint recovery failure for compacted changelog files
Purpose
Fixes checkpoint recovery failures when using precommit-compact functionality introduced in commit [flink] add coordinate and worker operator for small changelog files compaction (#4380).
Root Cause:
Compacted changelog files have two types of file names:
compacted-changelog-xxx$bid-len.cc-formatcompacted-changelog-xxx$bid-len-off-len2.cc-formatFake file names point to segments of real files but don't exist in the filesystem. The
checkFilesExistencemethod was directly checking these fake file paths, causing recovery failures.Solution:
CompactedChangelogPathResolverutility class to resolve fake file paths to real file pathsTableCommitImpl.checkFilesExistence()to resolve all paths before checking existencexxx$bid-len.cc-format): return original pathxxx$bid-len-off-len2.cc-format): resolve tobucket-bid/xxx$bid-len.cc-formatTests
CompactedChangelogPathResolverTestto verify path resolution logicAPI and Format
No changes to public API or storage format. This is an internal fix for file path resolution.
Documentation
No new features introduced. This is a bug fix for existing precommit-compact functionality.