TheR1sing3un commented on code in PR #17472:
URL: https://github.com/apache/hudi/pull/17472#discussion_r2587598605


##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/timeline/versioning/v2/LSMTimelineWriter.java:
##########
@@ -208,16 +209,19 @@ private void createManifestFile(HoodieLSMTimelineManifest 
manifest, int currentV
     int newVersion = currentVersion < 0 ? 1 : currentVersion + 1;
     // create manifest file
     final StoragePath manifestFilePath = 
LSMTimeline.getManifestFilePath(newVersion, archivePath);
-    metaClient.getStorage().createImmutableFileInPath(manifestFilePath, 
Option.of(HoodieInstantWriter.convertByteArrayToWriter(content)));
+    // create the manifest file with overwrite semantics, to handle the case 
like that:
+    // writer_1 creates the manifest file successfully, but fails before 
updating the version file,
+    // so we need to allow writer_2 to overwrite the manifest file wth the 
same version number.
+    FileIOUtils.createFileInPath(metaClient.getStorage(), manifestFilePath, 
Option.of(HoodieInstantWriter.convertByteArrayToWriter(content)));

Review Comment:
   > The method `createImmutableFileInPath` should keep the atomicity
   
   I agree that.
   
   But what I want to express is that the entire archive process should 
maintain atomicity, that is, creating archived data file, manifest and 
modifying version should maintain atomicity. 
   However, it is possible for us to fail at any step now. 
   For instance, if the failure occurs between creating the manifest and 
modifying the version, an invalid manifest file will remain on the file system, 
which will cause our retry to fail.
   What I want to solve is the atomicity problem in this process



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to