Chung-En Lee created HDDS-17869:
-----------------------------------

             Summary: S3 DeleteObjects on FSO bucket recursively deletes 
non-empty directories
                 Key: HDDS-17869
                 URL: https://issues.apache.org/jira/browse/HDDS-17869
             Project: Apache Ozone
          Issue Type: Bug
          Components: Ozone Manager, S3
    Affects Versions: 2.0.0, 2.1.0, 1.4.1, 2.2.2, 2.2.0
            Reporter: Chung-En Lee


On FSO buckets, S3 DeleteObjects with a directory key ({{dir/}}) deletes the 
whole subtree, including objects not listed in the request. A single 
DeleteObject on the same key is a no-op.

Regression from HDDS-10572 (1.4.1): multiDelete now calls {{deleteKeys}}, whose 
FSO handler {{OmKeysDeleteRequestWithFSO}} (written for recursive bucket 
deletion) has no emptiness check. 1.3.0 and 1.4.0 return {{Directory is not 
empty}} and keep the data; 1.4.1 through 2.2.2 delete it.

Impact: s3a with {{fs.s3a.directory.marker.retention=delete}} (default before 
Hadoop 3.4, used by Flink's S3 plugin) sends this request for parent markers 
after every write, wiping sibling data such as Flink checkpoints.

Repro on an FSO bucket:
{code}aws s3api put-object --bucket fso --key c/d2/file1.txt --body f
aws s3api put-object --bucket fso --key c/other.txt --body f
aws s3api delete-objects --bucket fso --delete 
'{"Objects":[{"Key":"c/d2/"},{"Key":"c/"}]}'
aws s3api list-objects-v2 --bucket fso --prefix c/
{code}
Expected: both files remain. Actual: both keys reported {{Deleted}}, and {{c/}} 
is empty.

Suggested fix: make multi-delete treat a non-empty directory as a no-op, like 
single DeleteObject.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to