It would be useful it we could enable the metadata_csum feature on the MDT/OST, so long as this is being properly handled by e2fsck. It would be a valuable addition to the "inode badness" metric that we currently have added to e2fsck. It isn't quite as useful by itself as the block checksums used by ZFS, since ZFS also has backup copies of all metadata blocks, but in some cases metadata_csum can still be valuable to detect corruption and avoid the use of corrupted metadata.
There are potentially conflicts between metadata_csum and other Lustre-specific ldiskfs features, such as dirdata inserting FIDs into the directory entries, so this must be reviewed and validated before the feature could be enabled by default.
- is related to
-
LU-17423 Upstream dir_data feature
-
- In Progress
-
