Shilun Fan created HDDS-15291:
---------------------------------
Summary: DN should fail unreferenced block deletion on file errors
Key: HDDS-15291
URL: https://issues.apache.org/jira/browse/HDDS-15291
Project: Apache Ozone
Issue Type: Improvement
Components: Ozone Datanode
Reporter: Shilun Fan
Assignee: Shilun Fan
Datanode may silently ignore failures while cleaning up unreferenced
block/chunk files during block deletion.
In KeyValueHandler.deleteUnreferenced, the current implementation lists chunk
files by prefix and deletes matching files with FileUtil.fullyDelete(file).
However, it does not handle two failure cases explicitly:
* File.list(...) may return null if the chunk directory cannot be listed.
* FileUtil.fullyDelete(file) returns false when deletion fails, but the result
is currently ignored.
As a result, DN may log that an unreferenced chunk/block was deleted even when
the file deletion actually failed. This can make cleanup failures harder to
detect and may leave stale unreferenced files on disk.
The proposed fix is to make deleteUnreferenced fail explicitly by throwing
IOException when the chunk directory cannot be listed or when file deletion
fails, and only log successful deletion after the file is actually removed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
