Predictive failure on a server drive: what it means and how to replace it safely

A predictive failure alert means the drive itself, through SMART, has reported that it is likely to fail. The drive is still working, and in a redundant array the data is still protected. That is the point of the alert: you can replace the drive on your schedule, while redundancy is intact, instead of after it fails. This guide covers how Dell and HPE report predicted failures and the replacement methods their documentation describes.

How the alert shows up

• Dell: the PERC controller checks drives for failure predictions and passes them to iDRAC, which logs an alert immediately (iDRAC9 user guide). PERC 12 also shows "A predictive failure is detected in the physical drive: <PD Name>" with the action "Replace the drive and ensure that the rebuild or copy-back operation is successful." Dell notes that PERC H355 and H350 controllers do not monitor predictive failures for non-RAID disks.
• HPE SR controllers: IML messages such as "Slot %1 Storage: Drive(s) at %2 status changed to predictive failure" and "Drive Array - S.M.A.R.T. Hard Drive(s) imminent failure". HPE describes the cause as "a drive predictive failure condition was detected. The drive needs to be replaced in the future."
• HPE MR controllers (Gen11): the IML event DrivePredictiveFailure, logged once per drive.
• HPE SSDs: an "SSD Smart Wear error" means the device is approaching its maximum write limit; the Gen11 troubleshooting guide's action is to replace the device.

HPE drive carrier LEDs differ between carrier types (SmartDrive, Basic, EDSFF), so check the LED table in the user guide for your server.

Step 1 — Dell: replace with Replace Member where possible

Replace Member copies data from the drive that is failing to a new drive while the virtual disk stays redundant. The PERC 12 guide gives this order for a SMART error on a drive in a redundant virtual disk:

1. Back up your data.
2. If Replace Member is supported and an empty slot is available, install the new drive in the empty slot and use Replace Member to copy to it, then verify the data against the backup.
3. If not, replace the affected drive with one of equal or higher capacity; an automatic rebuild starts. Verify the data afterwards.

In the PERC 12 HII utility the manual operation is in Virtual Disk Management: select the virtual disk, expand its physical disks, highlight the disk, press F2, select Replace, then Start, choose the replacement and select OK. The target must be a hot spare or an unconfigured disk without a foreign configuration, of the same type and the same or greater capacity. The PERC 11 guide gives a different sequence for redundant disks: back up, force the disk offline (a hot spare starts rebuilding if present), replace it, then perform Replace Member.

From iDRAC9, the RACADM guide documents the same copy as Replace Physical Disk. The source must be online and the destination in Ready state, of similar size and type:

racadm storage replacephysicaldisk:<Source PD FQDD> -dstpd <Destination PD FQDD>

To have this happen automatically, PERC 11 offers Main Menu > Controller Management > Advanced Controller Properties > Spare > Auto Replace Member on Predictive Failure. It needs a hot spare that qualifies as a rebuild disk; the drive with the SMART error is marked failed only after the copy completes, so the array does not go Degraded. PERC 12 lists separate HDD and SSD options for H975 and H965 series controllers. Physical replacement steps are in replace a failed disk on Dell PERC.

Step 2 — HPE: check the volume status, then replace

1. Confirm the logical drive or volume status is OK in iLO (System Information > Storage) or SSA. The IML action for a drive in a fault-tolerant volume is: verify the Volume Status is OK, then replace the drive.
2. For a drive that is not in a fault-tolerant volume, HPE's action is: back up all data, replace the drive, reconfigure the RAID, and restore from backup.
3. Replace the drive as described in replace a failed drive on Smart Array and watch the rebuild complete.

HPE SR controllers can also act on predicted failures automatically. In predictive spare activation mode, a spare activates when a member drive reports a predictive failure; data is copied to the spare while the volume is still healthy, and the failing drive is then marked failed and ready for removal. After you install the replacement, the controller restores data from the activated spare to the new drive. The SSA CLI syntax, with the guide's example slot:

controller slot=1 modify spareactivationmode=predictive

This needs an assigned spare; see HPE Smart Array spare drives.

Step 3 — Open a warranty case

A predictive failure is a valid reason to request a replacement. HPE's IML guidance is to verify the component is under warranty or contract and then create a support case. For Dell, collect a SupportAssist collection so the alert and drive details are attached to the case.

Frequently asked questions

Can I wait until the drive actually fails?

The drive is still working, but the vendors' guidance is to replace it: HPE's LED definition says "replace the drive as soon as possible". Replacing it while the volume is healthy avoids running degraded.

Should I pull the drive straight away?

Not in a non-redundant volume, and not before checking the volume status. In a RAID 0 volume, pulling the drive takes the data with it, so back up first.

Checklist

• Alert confirmed in iDRAC Lifecycle Log or iLO IML; drive slot and serial noted.
• Volume type checked: redundant, RAID 0 or non-RAID.
• Backup confirmed; volume status OK/Optimal before any replacement.
• Replace Member (Dell) or predictive spare activation (HPE) used where available.
• Rebuild or copy-back completed and volume back to Optimal/OK.
• Warranty case opened with logs attached.

Skyline can supply, configure and maintain Dell, HPE and Huawei servers under an annual maintenance contract (AMC). Contact us.