RAID 5 Capacity, Predictive Failure, and Online Expansion on an R760
An R760 is configured with a 16 x 2.5-inch front backplane, a PERC H755 front controller, two BOSS-N1 M.2 boot drives, and eight 3.84TB SAS SSDs. The workload is a PostgreSQL cluster with a 70/30 read/write mix. The customer created one RAID 5 virtual disk across seven SSDs and assigned the eighth SSD as a global hot spare. They expected about 26.8TB of usable capacity, but the operating system reports about 23.0TB. In the first month, the drive in slot 4 begins logging predictive-failure warnings in iDRAC, although it remains online. The customer wants no outage during an upcoming quarter-end close and asks whether the hot spare will take over automatically before the drive actually fails. They also plan to add two more 3.84TB SSDs next quarter to grow the same virtual disk.
1. Calculate the current usable capacity and explain the customer's 26.8TB expectation. What does the hot spare contribute, and what layout would actually deliver about 26.8TB usable with one spare?
Read the worked answer
RAID 5 usable capacity is the number of data drives multiplied by drive size. Eight SSDs are installed, but one is assigned as a hot spare, so only seven are members of the virtual disk. RAID 5 reserves one member's worth of capacity for parity, leaving six data drives: 6 x 3.84TB = 23.04TB, which matches the OS report before file-system overhead and decimal-versus-binary reporting. The customer's 26.8TB figure is 7 x 3.84TB, which is what an eight-drive RAID 5 with no hot spare would yield. The hot spare contributes zero addressable capacity; it is a standby that the controller can use for a rebuild. If the requirement is about 26.8TB usable plus one spare, the array needs eight data drives plus a ninth drive as the spare. With the current 16-bay chassis there is room, but the customer must choose between capacity and an immediate spare. For a quarter-end close with no outage tolerance, keeping the spare is reasonable, but it is not a substitute for backups.
2. The slot 4 drive is still online but flagged predictive failure. Describe the safest way to use the hot spare proactively, what happens after the original drive is replaced, and which controller settings you would verify.
Read the worked answer
Do not pull the slot 4 drive. Removing it forces the virtual disk into a degraded state and starts a parity rebuild, which is exactly the risk you are trying to avoid. The PERC supports a Replace Member Disk operation: in iDRAC, go to Storage > Controllers > PERC H755 > Virtual Disks, select the RAID 5 virtual disk, and choose Replace Member. Set the source to the slot 4 physical drive and the target to the global hot spare. The controller copies the data to the spare while the virtual disk stays optimal; this is a copy, not a rebuild from parity. When the copy completes, the hot spare becomes a regular member and slot 4 is marked as replaced or failed. Replace the physical drive, then check whether Auto Copyback is enabled so the data can return to slot 4 and the spare can return to standby. Verify Auto Rebuild, the hot-spare assignment (global versus dedicated), patrol-read schedule, and rebuild/copyback rate. Confirm the operation in the Lifecycle Log and with a SupportAssist collection.
3. The customer adds two 3.84TB SSDs next quarter. What must be true to expand the existing RAID 5 online, what capacity results, and what risks or follow-up steps apply?
Read the worked answer
Online expansion requires free, compatible drives of the same interface and type, and a controller that supports Reconfigure. The H755 supports adding drives to a RAID 5 virtual disk. Adding two 3.84TB SSDs to the current seven-member RAID 5 gives nine members, so usable capacity becomes (9 minus 1) x 3.84TB = 30.72TB, assuming both are added as data drives. If the customer also wants a hot spare, one of the new drives may need to be assigned as a spare, which reduces the gain. Before expanding, back up, confirm the chassis has free bays and the backplane is cabled, and check that the drives are Dell-qualified and not part of another virtual disk. Expansion triggers a background initialization or reconstruction that can run for hours and affect latency. After the controller expansion, grow the partition and file system in the OS; the controller operation alone does not enlarge the volume. Reassess fault tolerance: a nine-drive RAID 5 still tolerates only one failure, and rebuild time grows with member count.