| Freezing or snapshotting the OCR / voting (Grid Infrastructure) disk group |
Clusterware node evictions or full cluster outage |
Snapshot the database disk groups only (DATA, FRA, REDO). Exclude GRID / CRS / +OCR from the Protection Group and from any quiesce. |
| Missing a disk group (e.g. REDO) in the snapshot |
Inconsistent recovery point |
Always snapshot ALL database disk groups together. Verify Protection Group membership. |
| Wrong volume added due to a mis-read WWID/NGUID |
Snapshot misses a real DB volume or includes an unrelated one |
SCSI: strip the 3624a9370 prefix, uppercase, match to purevol list. NVMe: match the last 5 bytes of the NGUID to the volume serial. Never assume ASM disk names match array volume names. Keep the Step 1 mapping table. |
| Individual volume snapshots instead of a Protection Group |
Write-order inconsistency risk |
Always use Protection Group snapshots for atomic, write-order-consistent capture. |
| Mixed storage arrays in one ASM disk group |
Cannot capture atomically in a single Protection Group |
Keep each disk group on one array, or use the application-consistent workflow with one Protection Group per array. |
| Database left in backup mode or suspended after a failure |
Redo overhead or stalled I/O |
Use error traps so END BACKUP / RESUME always runs, even when the snapshot fails. |
| Defaulting to crash-consistent without checking constraints |
Skips the Oracle-consistent recovery point, or relies on crash recovery unnecessarily |
Confirm whether a quiesce window is achievable before choosing crash-consistent. Use the application-consistent workflow by default. |