By manufacturer
Dell PowerEdge HPE ProLiant Lenovo ThinkSystem Fujitsu Primergy Supermicro IBM System x / Power Acer AltosFailed RAID controller: replacement that preserves the array. No destructive rebuilds.
Hardware HBA / RAID cards (PERC, Smart Array, MegaRAID, Adaptec, ThinkSystem 9350/940) replaced with the configuration intact: foreign config re-import, no unwanted rebuild, BBU/Flash cache batteries handled correctly. Including after a controller failure on a live array.
Controller failed, array foreign. What NOT to do.
When a hardware RAID controller fails and is swapped for a new or replacement unit, the array on the disks is reported as foreign. At this point it is very easy to do damage:
- Do not run "Initialize": it destroys the existing configuration. Once the controller has been initialised, the array on the disks is unusable by standard means (it would take advanced logical recovery).
- Do not "Clear Foreign Config" before reading it: understand the disk order and slot mapping before touching anything.
- Do not run a "precautionary" rebuild with no actual degradation: it adds stress to the array at its most fragile moment.
Safe re-import, validation, return to production.
- Up-front documentation: physical disk order (photos), controller output, any predictive events.
- Foreign config re-import: import of the existing array, metadata integrity check, consistency check between the new controller and the old one (firmware, supported RAID level).
- BBU / Flash backup handling: the new controller's battery-backed cache must be healthy before write-back is enabled. Write-through policies if the BBU is not ready.
- Validation: full read pass across the array, I/O verification with no CRC errors, monitoring for 4-24h before the return to production.
The questions we get most often.
Can I replace a PERC H730 with an H740 or an H755?
Generally yes, but with checks: firmware compatibility with the installed disks, RAID level supported by the newer generation, cache settings to be reconfigured. On Dell PowerEdge, cross-compatibility is documented in the vendor matrix — we verify it before the job.
What happens to the data if the RAID controller fails hardware-wise?
The data stays on the disks: the controller is the arbiter, not the container. The risk is not losing the data to the controller fault itself — it is losing the data to the wrong operations performed on the new controller (initialize, clear config). That is why the first step is always hands off until the procedure is clear.
Hardware RAID vs software RAID: what do you recommend for new workloads?
For new workloads, increasingly often software RAID (mdadm on Linux, ZFS on BSD/Linux, Storage Spaces on Windows Server) offers advantages: independence from the controller vendor, more transparent recovery, often better performance with NVMe. On existing workloads with a healthy hardware RAID controller: there is no reason to change for the sake of changing. Moving to SDS is a separate project.