-
Question/Request
-
Resolution: Unresolved
-
Medium
-
None
-
None
-
None
-
Rocky Linux 9.4 (Blue Onyx)
5.14.0-427.31.1_lustre.el9.x86_64
-
3
-
9223372036854775807
Dear Whamcloud Staffs,
We encountered this strange error messages on our Lustre environment while end-users were trying to upload some files:
[Sat Aug 8 16:07:42 2026] LDISKFS-fs error (device sda): __ldiskfs_iget:5112: inode #3039385605: block 7960169: comm mdt00_002: invalid block [Sat Aug 8 16:07:42 2026] LDISKFS-fs error (device sda): __ldiskfs_iget:5112: inode #3039385602: block 28271: comm mdt00_002: invalid block [Sat Aug 8 16:07:42 2026] LDISKFS-fs error (device sda): __ldiskfs_iget:5112: inode #3039385605: block 7960169: comm mdt00_002: invalid block ...
We noted that not all files are affected.
The issue seems to be linked to the following:
LU-19121 ldiskfs/e2fsck mishandle fast symlink with external xattr inode - Whamcloud Community JIRA
We also successfully reproduced this issue on our lab environment with oversize xattr. We tried to fix the metadata with lfsck but it always fails to repair the two files with oversized xattr with the following messages:
[root@lustre-mdsvm2 ~]# lctl lfsck_start -M lustreFS-MDT0001 -A Started LFSCK on the device lustreFS-MDT0001: scrub layout namespace [root@lustre-mdsvm2 ~]# lctl get_param osd-ldiskfs.lustreFS-MDT0001.oi_scrub osd-ldiskfs.lustreFS-MDT0001.oi_scrub=name: OI_scrub magic: 0x4c5fe253 oi_files: 64 status: stopped flags: param: time_since_last_completed: N/A time_since_latest_start: 108 seconds time_since_last_checkpoint: 108 seconds latest_start_position: 12 last_checkpoint_position: 3977402881 first_failure_position: 3039385602 checked: 97 updated: 0 failed: 2 prior_updated: 0 noscrub: 11 igif: 1 success_count: 0 run_time: 0 seconds average_speed: 97 objects/sec real_time_speed: N/A current_position: N/A lf_scanned: 1 lf_repaired: 0 lf_failed: 0
We do not have full understandings of the bug coverage and would like to ask the following questions:
- Whether it is possible to mitigate this "invalid block" issue by upgrading to 2.17.0
- Whether the affected files can be recovered, or they are completely lost.
- Is oversize xattr the only possible reason behind this invalid block error ? We afraid that if we upgrade Lustre to 2.17.0 and if it does not completely fix the invalid block metadata, we won't be able to remount the Lustre service and lose all accesses to Lustre
I would appreciate if anyone can shed lights on this matter.
Best regards,
Minh