Before running a repair, you must find the identifier for the backed device (NAA ID) or the specific partition. Log into the ESXi shell via SSH.
| ID | Requirement | |----|--------------| | F5 | Support (pre-check). | | F6 | Provide granular repair options: | | - Repair FBM inconsistencies | | - Rebuild FDC from secondary copy (if available) | | - Replay/recover journal entries | | - Fix orphaned file descriptors | | - Remove stale lock entries | | F7 | Safe mode: Only repair metadata; never touch user data blocks. | | F8 | Force mode: For severely damaged volumes (skip some checks). | | F9 | Automatic file system checkpoints (if underlying storage supports snapshots). | vmfs repair file system
This involves manipulating partition tables. Incorrect use results in data loss. Before running a repair, you must find the
esxcli storage vmfs snapshot list
vmkfstools -e /vmfs/devices/disks/<DEVICE_ID> | | F6 | Provide granular repair options:
| Edge Case | Handling Strategy | |------------|-------------------| | Multiple ESXi hosts mounted the same LUN | Verify heartbeat; pause repair if active heartbeats > 0. | | FBM and Journal both corrupt | Attempt journal replay first; if fails, rebuild FBM from block allocation tables. | | Disk has bad sectors (hardware failure) | Warn user → suggest vmfs-repair --skip-bad-sectors + export list of affected files. | | VMFS version mismatch (e.g., upgraded from 5 to 6) | Use backward-compatible repair logic. |
| ID | Requirement | |----|--------------| | F13 | of critical metadata regions before any write operation. | | F14 | Host quiesce detection – block repair if VMs are actively writing (unless force). | | F15 | Validation after repair: Run consistency check immediately after repair. |