igalshilman commented on a change in pull request #7161:
[FLINK-10963][fs-connector, s3] Cleanup tmp S3 objects uploaded as backups of
in-progress files.
URL: https://github.com/apache/flink/pull/7161#discussion_r237019573
##########
File path:
flink-filesystems/flink-s3-fs-base/src/main/java/org/apache/flink/fs/s3/common/writer/RecoverableMultiPartUploadImpl.java
##########
@@ -209,6 +204,10 @@ static String incompleteObjectNamePrefix(String
objectName) {
return parent + (child.isEmpty() ? "" : '_') + child + "_tmp_";
}
+ private String createIncompletePartObjectName() {
+ return namePrefixForTempObjects + UUID.randomUUID().toString();
Review comment:
I'm wondering here why do we use `UUID` here, would that be possible to
create the name deterministically as a function of the current jobid, attempt
id, checkpoint id and the `namePrefixForTempObjects`?
(unless they are not accessible at that level of abstraction)
Having the name deterministic seems like a good building block to address
the limitation described in this PR description.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
