Account
Key takeaways
- Identify the failed member from the controller inventory and the amber LED before touching any drive.
- Match the interface: SAS controllers accept SATA drives on most backplanes, but a SATA controller never accepts a SAS drive.
- The replacement must be at least the member's capacity; extra capacity is usually left unused, and a smaller drive is rejected.
- Match sector format (512n, 512e or 4Kn), spindle class or SSD endurance class, and use OEM firmware where the platform expects it.
- Rebuild time depends on capacity, RAID level and load, and a RAID 5 array has no redundancy until the rebuild finishes.
How do you identify the failed drive?
Start with the controller, not the LED. iDRAC (Dell PERC), iLO or Smart Storage Administrator (HPE Smart Array) and the controller's own utility list each physical disk with slot, state, model, serial number, capacity, interface, sector size and firmware. Record everything the inventory shows about the failed member; it is your matching specification.
Then confirm at the chassis. Most carriers have a status LED that turns amber or orange on failure; the exact pattern varies by vendor, so read your model's LED legend. Most controllers can blink a chosen slot's LED on demand; use it, because pulling the wrong drive from a degraded RAID 5 array takes the whole array offline.
How do you match the interface: SAS, SATA or NL-SAS?
SAS and SATA drives look alike, but a SAS drive has a second port and speaks a different protocol. A SAS controller and backplane accept SATA drives on most servers; a SATA-only controller cannot use a SAS drive. NL-SAS (nearline SAS) is a 7,200 rpm drive with a SAS interface.
Within one virtual disk, keep the interface consistent. Most controllers refuse to add a SATA drive to a SAS array or the reverse, and those that allow it warn you. NVMe drives connect over PCIe, not through a SAS controller, and are not a substitute for a SAS or SATA member.
Does the replacement drive need the same capacity?
At least the same, and usually exactly the same. The controller stores the array layout in terms of each member's usable sector count, so a replacement with fewer sectors is rejected. A larger replacement is accepted on most controllers, but the array only uses the size it already has; the extra space sits idle unless every member is later replaced and the array expanded.
Drives sold at the same capacity by different manufacturers can differ by a few sectors. Many controllers round capacity down to a common boundary (Dell calls this coercion) so a marginally smaller replacement still fits, but not all do, which is why matching the original OEM part number is the safe path.
What is sector format and why does it have to match?
Sector format is the logical sector size the drive presents to the controller and the physical sector size on the media.
| Format | Logical sector | Physical sector | Notes |
|---|---|---|---|
| 512n (native) | 512 bytes | 512 bytes | Older and smaller drives |
| 512e (emulated) | 512 bytes | 4,096 bytes | Most current hard drives; looks like 512n to the controller |
| 4Kn (native) | 4,096 bytes | 4,096 bytes | Needs controller, OS and UEFI boot support for 4K logical sectors |
A virtual disk is built on one logical sector size. Most controllers will not add a 4Kn drive to a 512-byte array or a 512n or 512e drive to a 4Kn array. 512n and 512e can usually coexist because the controller sees 512-byte sectors on both. Buy the format the inventory shows. The spec decoder covers reading 512e and 4Kn from a label.
Do spindle speed and SSD endurance have to match?
For hard drives, match the spindle class: 7,200 rpm (NL-SAS and SATA), 10,000 rpm or 15,000 rpm. Most controllers allow mixed spindle speeds in one virtual disk, but the array performs like its slowest member. Match the form factor too: a 2.5-inch drive in a 3.5-inch bay needs an adapter carrier.
For SSDs, match the endurance class. Datasheets express endurance as drive writes per day (DWPD), and OEMs group drives as read-intensive, mixed-use or write-intensive. A read-intensive drive standing in for a write-intensive member wears faster than the rest of the array. Match the interface (SAS or SATA SSD) as well.
Does the drive need OEM firmware?
Dell and HPE ship drives with their own firmware. Current Dell PERC generations accept non-Dell drives but flag them in iDRAC as uncertified and exclude them from Dell firmware updates. HPE Smart Array controllers expect HPE drives with digitally signed firmware; other drives generally function but appear as unsupported in iLO and Smart Storage Administrator, and HPE may decline support for the array.
The practical rule: in a supported OEM server, replace an OEM drive with the same OEM part number so monitoring, firmware updates and support keep working. In an EOL fleet on third-party maintenance, a non-OEM drive of the correct specification is often acceptable; our EOL and EOSL guide describes how those fleets are run.
What is the hot-swap procedure?
Hot-swap means replacing the drive with the server running. The steps are the same on most controllers.
- Confirm the failed slot in the controller inventory and blink its LED. Confirm a recent backup exists.
- Check the replacement against the inventory: interface, capacity, sector format, class and form factor.
- Release the latch, pull the failed drive out a short distance, wait a few seconds for it to spin down, then remove it.
- If the replacement shipped bare, fit it into the original carrier.
- Insert the new drive until the latch closes fully. The controller should detect it within seconds.
- If automatic rebuild is enabled, the rebuild starts on its own. If not, assign the drive as a replacement or hot spare in the controller utility. If it shows as foreign, clear the foreign configuration first.
- Monitor until the virtual disk reports optimal.
Rebuild time depends on capacity, RAID level, the controller's rebuild-rate setting and how busy the array is; large hard drives take hours to days. During a RAID 5 rebuild there is no redundancy left, so a second failure loses the array; RAID 6 survives one more.
What about the tray, spares and the quote?
Drives mount in a carrier (tray or caddy) specific to the chassis generation and form factor: HPE's SmartCarrier for Gen8 and later differs from earlier ProLiant trays, Dell's carriers changed between generations, and 2.5-inch, 3.5-inch and NVMe carriers all differ. Reuse the failed drive's carrier when the replacement ships bare. Our drive tray guide covers which carrier fits which chassis.
Keep a cold spare per drive model, in the correct carrier, with the same firmware as the installed members, and set a hot spare wherever a bay is free. Buy spares while the model is still easy to source.
For a quote, send the server model, the controller model (PERC H730P or Smart Array P408i-a, for example), the existing member's part number, capacity, interface, class and sector format as the controller inventory shows them, whether you need a carrier, and the quantity. The part number guide shows where each value lives.
Memory Clearance sells new and refurbished server drives and quotes per part. Quotes are answered weekdays 8:00 am to 5:00 pm PST. The condition shown on a listing is the condition you receive; warranty terms are on the warranty page.
Frequently asked questions
Can I use a larger drive to replace a failed member?
On most controllers, yes. The controller accepts the drive and uses only the capacity the array already has; the extra space stays unused. It is a reasonable option when the exact size is no longer available. A drive with fewer usable sectors than the member is rejected, so never go smaller.
Will a non-Dell drive work in a Dell PERC array?
On current PERC generations it generally works, but iDRAC flags it as uncertified and Dell firmware updates skip it. For a supported production server, match the Dell part number; for an EOL fleet on third-party maintenance, a correct-spec drive of the same class is often acceptable.
How long does a RAID rebuild take?
It depends on the drive capacity, the RAID level, the controller's rebuild priority setting and how busy the array is. A multi-terabyte hard drive in a busy RAID 5 or RAID 6 array can take a day or more; an SSD array is usually much faster. Avoid heavy workloads during the rebuild.
Can I replace a failed hard drive with an SSD in the same array?
Most controllers refuse to mix SSDs and hard drives within one virtual disk, and the ones that allow it give you an array that behaves like its slowest member. Replace like with like. If you want SSD performance, build a new virtual disk on SSDs and migrate the data.