RAID Cache Battery Failure: Why Write-Back Gets Disabled and What to Replace
Your array didn't lose a disk — the controller quietly disabled write-back caching. How to confirm a dead BBU or supercap and match the exact replacement.
When a RAID array that ran fine for years suddenly writes like a single laptop drive, the usual culprit isn't a failed disk. It's a failed cache battery. The controller noticed, disabled write-back caching to protect your data, logged one line in its event log, and said nothing else. The operating system sees nothing wrong — just a volume that got slow.
Why the controller disables write-back on its own
Write-back caching lets the controller acknowledge a write as soon as it lands in onboard DRAM, then flush to disk on its own schedule. That is where hardware RAID gets most of its write performance. The catch: if power fails with unflushed data in cache, those writes are gone, and on parity RAID the stripe they belonged to can be left inconsistent.
The battery backup unit (BBU) — or on newer flash-backed designs, a supercapacitor that powers a dump of cache contents to onboard NAND — exists to close that window. When the controller can no longer trust the module, whether it has failed outright, degraded below a charge threshold, or (on older lithium-ion packs) entered a periodic learn cycle, it reverts to write-through: every write must reach the disks before it is acknowledged. This is deliberate, correct behavior. It is also nearly silent unless you monitor the controller itself.
What write-through does to your array
Small random writes suffer most. On RAID 5 and RAID 6, each small write becomes a read-modify-write parity sequence the cache previously absorbed and coalesced, so sustained small-write throughput on spinning disks can drop to a small fraction of its cached level. Databases, mail servers, and VM datastores — anything that syncs constantly — feel it first.
Confirming takes one command:
- Dell PERC: perccli /c0/bbu show all, or check battery status in iDRAC/OMSA
- HPE Smart Array: ssacli ctrl slot=0 show detail and read the battery/capacitor status lines
- Broadcom/LSI MegaRAID: storcli /c0/bbu show all for batteries, storcli /c0/cv show all for CacheVault
In the same output, check the virtual drive's cache policy. WriteBack downgraded to WriteThrough means the controller has already made its decision.
Matching the replacement to the controller
Cache protection modules are matched to the controller family, not the server model. The broad map:
| Controller family | Cache protection |
|---|---|
| Dell PERC 6/i, H700 | Li-ion BBU with periodic learn cycles |
| Dell PERC H730/H740P | Battery that powers a flush to non-volatile flash |
| HPE Smart Array P410 (G6/G7) | BBWC battery pack |
| HPE P420/P440 (Gen8/Gen9) | FBWC capacitor pack — no learn cycles |
| HPE Gen9/Gen10 (P440ar, P408i) | Shared Smart Storage Battery |
| MegaRAID 9260/9271 | iBBU07/08/09 battery, per Broadcom's compatibility list |
| MegaRAID 9361/9380 | CacheVault supercapacitor kit |
The safest source of truth is the part number printed on the failed module itself, since cable lengths and connectors vary even within a family. If the controller reports memory or firmware faults alongside the battery fault, price a replacement card too — browse our RAID controllers for both current and end-of-life generations, and our replacement batteries for BBUs and capacitor kits. These parts commonly fail on servers long out of production, which is exactly the hardware we keep in stock, backed by a 3-year warranty.
Hot-replacement caveats
- Check where the module actually lives. Many BBUs sit on a remote tray connected by cable and can be swapped with the chassis open; others mount on the card under an airflow shroud and effectively require downtime. HPE's shared Smart Storage Battery generally requires a shutdown.
- Write-back does not return the moment you plug in a new pack. A lithium-ion BBU must charge, and often complete an initial learn cycle, before the controller re-enables write-back — plan for hours. Supercapacitor kits recharge far faster.
- Never remove a module while the controller reports dirty or preserved cache after a power event. Let it flush first.
- Resist forcing write-back without a working battery ("Always Write Back" / "Force WB"). Unless the server sits on a UPS with tested graceful shutdown, one power cut loses whatever was in cache.
Not sure which module your card takes? Send us the controller model or the part number off the old battery and our sales team will match the exact replacement free of charge. In-stock parts ordered before 12 PM EST ship the same day.
Frequently Asked Questions
Why did my RAID controller disable write-back caching?
The controller detected that its cache battery or capacitor module failed, degraded below a charge threshold, or entered a learn cycle. Since it can no longer guarantee cached writes survive a power loss, it switches to write-through mode automatically — no OS-level alert is raised.
Can I force write-back without a working battery?
Most controllers offer a force write-back option, but a power loss or hard crash will discard everything in cache and can corrupt parity stripes. Only consider it on a UPS with tested graceful shutdown, and treat it as a stopgap until the battery is replaced.
Do flash-backed (FBWC/CacheVault) modules fail like batteries?
Yes. Supercapacitors age and lose capacitance, and the controller disables write-back the same way when it detects a degraded pack. They do avoid learn cycles and recharge much faster than lithium-ion BBUs after replacement.
Will write-back return immediately after I install a new battery?
No. A new lithium-ion BBU must charge and often complete an initial learn cycle before the controller re-enables write-back, which can take hours. Capacitor-based kits typically restore write-back within minutes of charging.