PVE 8.2 on Dell R420

[afede]

Hi,
I tried to install PVE 8.2 on a server Dell R420 with PERC 310 Mini integrated RAID controller. The RAID controller seems to work properly (booting a Ubuntu 22.04 live key shows me the underlying disks exposed by it) but when I boot up the pve installer it sais "No hard disk found".
Jumping into a shell (ctrl+alt+f3) and issuing "fdisk -l" shows me no disks (except the pve live key).
DMESG shows me only this messages:

batou kernel: DMAR: DRHD: handling fault status reg 3
batou kernel: DMAR: [DMA Read] Request device [05:00.0] fault addr fffe0000 [fault reason 06] PTE Read access is not set
batou kernel: DMAR: DRHD: handling fault status reg 3
batou kernel: DMAR: [DMA Read] Request device [05:00.0] fault addr fffe0000 [fault reason 06] PTE Read access is not set
batou kernel: DMAR: DRHD: handling fault status reg 3
batou kernel: DMAR: [DMA Read] Request device [05:00.0] fault addr fffe0000 [fault reason 06] PTE Read access is not set
batou kernel: DMAR: DRHD: handling fault status reg 3

I gave up installing a 7.4 version (which works correctly!)...

What I want to know you is that upgrading this server from PVE 7 (with kernel 5.x) to PVE 8 (with kernel 6.x) destroyed my working installation.
During the boot phase I can see grub, but when I try to boot the new kernel it says that it doesn't have the root disk.



Now my server is running PVE 8 on a 5.x kernel, but what do you think about this story?
Can it be a a bug of the megaraid driver under the newest kernel?

 

[Fantu]

With kernel 6.8 I also had issue with dell server for megaraid driver issue after upgrade with it as default and I workaround it using 6.5 kernel with

proxmox-boot-tool kernel pin <old-kernel-version>

I have not checked if is solved in latest kernel revision.
For new installation with 8.2 iso that have this kernel as default I don't know an easy/fast way to install older kernel instead, I think is easier/faster use 8.1 iso and after update the system, and if 6.8 still have megaraid issue pin to 6.5

 

[afede]

I upgraded yesterday to the latest kernel (proxmox-kernel-6.8.12-1). This one still has this problem.
Temporarily resolved pinning the old kernel with the command you said.

 

[waltar]

WIth T620 include H710 works fine from 7.x until 8.2.4 with 6.8.12-1 (our T330's are out-of-business yet so I cannot say about them).

 

[christianh32]

I'm running into the same issue as the OP. Dell R420/H310 mini. Works fine in 8.1 (6.5.13-6-pve), but if I upgrade to 8.2 (6.8.12-1-pve) or attempt a fresh install of 8.2 I get "No hard disk found". Pinning to 6.5.13-6-pve works, but I would love to get to apply a fix rather than a work around before creating VMs on it. Has anyone found a fix?

 

[waltar]

Get used H710 for a few bugs and enjoy function and performance win

 

[user73937393]

This problem seems similar to a problem I ran into, albeit on an HP ML310e. I'd suggest bypassing the RAID controller altogether and just use ZFS.

 

[christianh32]

Thanks all! It does indeed seem very similar. I was still hoping to still utilize the H310 RAID controller, but it does not appear it will work well. The R420 is a test host I was using prior switching from VMware. I know the hardware is on the older side and I might not experience these issues with new hardware. If someone does come across a fix I would love to attempt to apply it!

 

[jdancer]

Thanks all! It does indeed seem very similar. I was still hoping to still utilize the H310 RAID controller, but it does not appear it will work well. The R420 is a test host I was using prior switching from VMware. I know the hardware is on the older side and I might not experience these issues with new hardware. If someone does come across a fix I would love to attempt to apply it!

Way better off flashing the H310 to IT-mode and using ZFS RAID. H310 RAID by default has very low queues which leads to terrible IOPS. Flashing it to IT-mode gets you the max queues. Flash using this guide https://fohdeesha.com/docs/perc.html

I have a fleet of 12th-gen Dells PERCs flashed to IT-mode in production with no issues using ZFS and Ceph.

 

[christianh32]

I have not tried to flash the PERC, but not opposed to do so especially if its supported. Thanks for the guidance!