-
Bug
-
Resolution: Unresolved
-
Medium
-
Lustre 2.18.0
-
3
-
9223372036854775807
Newly formatted OSTs return an e2fsck error about unattached inode 88 repeatedly:
e2fsck 1.47.3-wc2 (11-Nov-2025) Pass 1: Checking inodes, blocks, and sizes Pass 2: Checking directory structure Pass 3: Checking directory connectivity Pass 4: Checking reference counts Unattached inode 88 Connect to /lost+found? no Pass 5: Checking group summary information myfs-OST0003: ********** WARNING: Filesystem still has errors ********** myfs-OST0003: 629/4294160640 files (0.2% non-contiguous), 537388688/4879024128 blocks
This inode is supposed to point at the failure_domain file which was added in LU-19066.
Running debugfs shows:
# debugfs -c -R "ls -l /" /dev/nvme1n10 | cat
debugfs 1.47.3-wc2 (11-Nov-2025)
2 40755 (2) 0 0 4096 7-Oct-2026 16:49 .
2 40755 (2) 0 0 4096 7-Oct-2026 16:49 ..
87 100644 (1) 0 0 0 1-Jan-1970 00:00 health_check
89 100644 (1) 0 0 8192 1-Jan-1970 00:00 nodemap
and health_check was created immediately before failure_domain in ofd_fs_setup():
lu_local_obj_fid(&info->fti_fid, OFD_HEALTH_CHECK_OID);
memset(&info->fti_attr, 0, sizeof(info->fti_attr));
info->fti_attr.la_valid = LA_MODE;
info->fti_attr.la_mode = S_IFREG | 0644;
info->fti_dof.dof_type = dt_mode_to_dft(S_IFREG);
fo = dt_find_or_create(env, ofd->ofd_osd, &info->fti_fid,
&info->fti_dof, &info->fti_attr);
if (IS_ERR(fo))
GOTO(out_seqs, rc = PTR_ERR(fo));
ofd->ofd_health_check_file = fo;
lu_local_obj_fid(&info->fti_fid, OFD_FAILURE_DOMAIN_OID);
fd = dt_find_or_create(env, ofd->ofd_osd, &info->fti_fid,
&info->fti_dof, &info->fti_attr);
There is no entry for failure_domain in osd_lf_maps[] in osd_scrub.c since no name is defined for failure_domain file. e2fsck pass 4 name linkage treats it as disconnected and will reconnect it in lost+found, which might accidentally be deleted by an admin at some point in the future, and e2fsck errors are always a concern for admins so they should be avoided.
