I installed a Startech PEX8M2E2 Dual M.2 PCIe SSD adapter card (recommended here) in an HP Microserver Gen10 Plus to use 2 M-Key SSDs.
After some time Proxmox reboots.
Following error is logged before the reboots:
Feb 23 08:21:48 pve3 kernel: NMI: PCI system error (SERR) for reason b0 on CPU 0.
Feb 23 08:21:48 pve3 kernel: Dazed and confused, but trying to continue
Feb 23 08:21:48 pve3 kernel: pcieport 0000:08:0c.0: can't change power state from D3cold to D0 (config space inaccessible)
Feb 23 08:21:48 pve3 kernel: pcieport 0000:08:04.0: can't change power state from D3cold to D0 (config space inaccessible)
I already updated Proxmox and the System ROM, changed the power management settings, used different kernels and enabled/disabled bifurcation without success.
I also asked HPE and Startech support (no solution yet) but maybe the Proxmox users here have an idea?
I’ve used about 5-6 Startech Dual NVME Adapters (last was also PEX8M2E2, but also used other models from Startech), alwayys on HP Servers (Different Models), never had any issues so far.
The oldest is now running for nearly two years now…
Maybe also important: What disks are you using there?
Any wear shown on the disks? (I had a couple of Samsung (970?) with faulty Disk-BIOS, the BIOS was updated, which fixed the issue. The wear: 35 % in 2 weeks!
I actually have the same HPE Microserver Gen10 Plus at home, I’m also following the same dok to use better CPU / More RAM. But this one I didn’t use the NVME, as I had 2 TB SSDs (new) lying around, also 2 x 128 GB SSDs as system…
And I had spare 2.5"->3.5" Adapters for the HPE Microserver…
I need the PCI for increased Networking, I’m using a 2x2.5 GBE NIC from Delock. The box is working well with 32 GB RAM, will upgrade to 64 GB RAM soon.
My 2 cents
Andy
PS:
You did read this (From said report by serverthehome.com you included…)
These PCIe switches use power, so we generally suggest avoiding them given how power-constrained the MSG10+ is…
This CAN easily be an issue, these boxes allow really max 180W all in all!
I’ve never handled these kinds of problems, but I’ll hope to find a way to work around the problem.
If pcieport 0000:08:0c.0 is where the adapter is connected you may try to disable pcie power management with pcie_port_pm=off on boot args.
But I fear Andy is correct, you have power supply problems.
I am curious because I just tossed two months ago a ssd nvme with PCIE 4X card in my old proxmox (10years old, pcie 2.0) and I must admit I should have done it before, really fast
There are options using PCIe switches such as the StarTech PEX8M2E2 adapter which offers two drive capability in a single PCIe slot without requiring bifurcation support.