Has anyone experienced a high failure rate when updating firmware on MR216i-A or MR216i-P HBAs in HPE DL380 Gen10 Plus servers? We have around a 80% chance of losing a card during a firmware upgrade, and we have already replaced at least nine. The servers are covered by support, but repeatedly swapping hardware is becoming frustrating. HPE documents this as a rare issue, though our experience has been very different. We have 15 hosts with three HBAs per host and would like to understand whether this is a known problem or if there is a safer way to perform the updates.
3 Answers
Are those nine failures all from one server or spread across the fleet? With 15 hosts and three HBAs per host, it would be worth tracking the exact server model revision, card part number, current and target firmware, update method, boot mode, and whether the system is rebooted between stages. That pattern sounds unusual enough that HPE support should investigate it as a repeatable hardware or firmware problem rather than treating each replacement as an isolated failure.
I haven’t seen this happen with these cards. We’ve used the same underlying LSI/Broadcom hardware, including OEM-branded versions, for years and have successfully completed firmware updates. One card was temporarily soft-bricked after a kernel panic interrupted the flash, but its recovery mode brought it back. Another needed to be flashed twice before it initialized properly under UEFI. That’s very different from permanently losing most cards during normal upgrades, so I’d suspect a platform-, firmware-package-, or procedure-specific issue rather than an inherent flaw in the silicon.
The cache or battery-backed modules are a separate reliability concern. We’ve had several of those fail, sometimes with the server sounding an alarm even though everything else was operating normally, but the HBA itself did not permanently fail during firmware updates. Make sure the update package is specifically intended for the HPE card, avoid interruptions during flashing, and use the documented recovery process if a card only appears bricked. A normal, uninterrupted update should not routinely destroy these adapters.
They’re spread across the environment. We have 15 hosts with three HBAs in each, and the failure rate has remained consistently high during upgrades.