An HP ProLiant DL380 G5 was restarted after an RPC-related problem, but it now powers on without displaying POST, BIOS, or any video output. There is also no hard-drive activity. Power was disconnected and drained, the RAM was reseated, the CMOS battery was cleared and reseated, and the RAID controller battery and hard drives were reseated. The fans and DIMM status lights turn green, but the fans remain loud and the system never progresses beyond startup. What should be checked next?
4 Answers
A permanently loud fan with no POST often means the system is stuck very early in hardware initialization rather than waiting on the disks. Remove the RAID controller, drives, and other PCIe cards, then try a minimal configuration and inspect every internal status LED. If it still produces no video or diagnostic code, the motherboard, CPU, VRM, or power subsystem may have failed.
The processor voltage-regulator module is worth investigating. A failed VRM for the first CPU can prevent POST and may also affect the second processor. If the server has two processors and matching VRMs, testing with one CPU and its known-good VRM—or temporarily swapping the VRMs if they are identical—can help isolate the fault. Replace the aging server as soon as possible if it comes back online.
This machine is still hosting the primary domain controller, and the replacement backup controller has not finished replicating yet, so the situation is urgent. I have not configured the management interface on this server, but I will check the LEDs and try a minimal-hardware boot before assuming the server is beyond recovery.
At this age, a hardware failure is very plausible. Check the system status LEDs for a specific fault code and see whether the integrated management controller is reachable, since its event log may identify a failed component. If there is no useful diagnostic information, test with the minimum hardware needed to POST: one CPU, the required VRM, one memory module, and no drives or expansion cards.
Prioritize getting another domain controller online from any available backup or recovery source rather than spending too long repairing this twenty-year-old system. If it does boot, treat that as an opportunity to complete replication and migrate the workload immediately.