I joined the plenty of people ordering a H12SSL-i + 7302p from ebay to replace my old dual X5650. Also found 4 sticks of MTA36ASF8G72PZ-3G2B2UI.
Everything arrived a few days ago and I mounted everything and connected a VGA monitor and a Ubuntu livecd on a Samsung SSD via USB->SATA adapter.
It ran memtest86+ for 12 hours, 2½passes, no issues. After that I connected 2x 250GB SATA SSD:s and installed ubuntu on a ZFS mirror, using rEFInd + ZFSBootMenu to boot it.
It booted fine, so I ran prime95 for 12 hours to check the CPU. No issues. So I moved the system closer to my desk and connected a network cable to the IPMI port and a cable to the first interface to start to migrate things. No monitor, but IPMI got an IP via DHCP, and after figuring out that my firefox security settings gave "Session timed out"-errors I logged in with chromium instead. The dashboard didn't give all sensor data, so I updated the BMC with BMC_H12AST2500-ROT-2201MS-GPIO01_20220524_01.01.06_STDsp.zip - but skipped BIOS update since it was recommended against by supermicro, unless I had some issues.
After that the sensor data was displayed just fine. I configured an IP-address in linux via IPMI, and then used SSH to connect and continue the configuration of libvirt and so on. After that I shut the power off and moved the Optane 900p, 1080TI GPU and two m1015-equivalents from the old server to the new and connected the monitor again, since a VM on the old server was acting DHCP server.
No image.
Of course, BIOS must be set to external GPU first, so I grab an HDMI-cable and connect the monitor to the 1080TI instead.
Much slower boot than before, but I get an image, BIOS flashes by and rEFInd starts, and goes:
---8<---
Starting vmlinuz.old-2.1.0_1
Using load options 'initrd=EFI\zbm\initramfs-2.1.0_1.img
EFI stub: Loaded initrd from command line option
_
---8<---
And then the keyboard goes black and it freezes. So I reboot, check the VGA output priority setting, and it shows as internal first. Odd. Maybe it's showing a default without being written to CMOS? So I switch it to external first, save & reboot, then enter bios again and set it to internal first. Same thing.
So I enter BIOS again, and set a static IP & netmask, no gateway, to IPMI to check out what it's saying, and reboot again. Time passes, and the IP never starts to reply to ping requests. Maybe the gateway was a required setting, even if it's not being used? Back into bios, and now it says:
BMC Firmware Revision: Unknown
IPMI STATUS: Not Working
Now I'm thinking maybe BIOS has to be updated in tandem with the BMC, but I didn't read anything about it, and the BMC didn't say anything about when i updated it via IPMI.
I cut power for all hardware changes. First removing all addon-cards, so I should be back in a working configuration, but no image via VGA. I remove all but 1 memory stick, same thing. Try another stick in another slot, same thing. So I add the 1080TI back again, and I get an image over HDMI.
I remove it again, clear CMOS, and switch to another battery while I'm at it. No changes.
I try booting the Ubuntu livecd over USB again. It loads GRUB, but hangs as soon as it starts to load the kernel, powering off the keyboard just as before.
Maybe this warrants a BIOS update, so I download latest BIOS_H12SSL-1B95_20220414_2.4_STDsp.zip and copy the contents to the EFI-partition of one of the linux drives I installed on earlier, and plug this in the USB->SATA adapter. Reading the help output for SUM.efi I realize it wants the IPMI password to flash a new BIOS, and thus as expected, it fails:
---8<---
FS0:\EFI\> flash.nsh BIOS_H12SSL-1B95_20220414_2.4_STDsp.bin ADMIN {{PASSWORD}}
[copyright header]
<<<ERROR>>>>
ExitCode = 101
Description = Driver input/output control failed
Program Error Code = 17.2
Error message: Send IPMI failed
---8<---
So, BMC seems broken, and I can't boot anything past UEFI-stuff anymore. Any Ideas?
No cratered IC as in the thread about "BMC Initiating", should I add pictures of the board?
I tried 2 other GPUs and another PSU too, just in case, no difference.
Everything arrived a few days ago and I mounted everything and connected a VGA monitor and a Ubuntu livecd on a Samsung SSD via USB->SATA adapter.
It ran memtest86+ for 12 hours, 2½passes, no issues. After that I connected 2x 250GB SATA SSD:s and installed ubuntu on a ZFS mirror, using rEFInd + ZFSBootMenu to boot it.
It booted fine, so I ran prime95 for 12 hours to check the CPU. No issues. So I moved the system closer to my desk and connected a network cable to the IPMI port and a cable to the first interface to start to migrate things. No monitor, but IPMI got an IP via DHCP, and after figuring out that my firefox security settings gave "Session timed out"-errors I logged in with chromium instead. The dashboard didn't give all sensor data, so I updated the BMC with BMC_H12AST2500-ROT-2201MS-GPIO01_20220524_01.01.06_STDsp.zip - but skipped BIOS update since it was recommended against by supermicro, unless I had some issues.
After that the sensor data was displayed just fine. I configured an IP-address in linux via IPMI, and then used SSH to connect and continue the configuration of libvirt and so on. After that I shut the power off and moved the Optane 900p, 1080TI GPU and two m1015-equivalents from the old server to the new and connected the monitor again, since a VM on the old server was acting DHCP server.
No image.
Of course, BIOS must be set to external GPU first, so I grab an HDMI-cable and connect the monitor to the 1080TI instead.
Much slower boot than before, but I get an image, BIOS flashes by and rEFInd starts, and goes:
---8<---
Starting vmlinuz.old-2.1.0_1
Using load options 'initrd=EFI\zbm\initramfs-2.1.0_1.img
EFI stub: Loaded initrd from command line option
_
---8<---
And then the keyboard goes black and it freezes. So I reboot, check the VGA output priority setting, and it shows as internal first. Odd. Maybe it's showing a default without being written to CMOS? So I switch it to external first, save & reboot, then enter bios again and set it to internal first. Same thing.
So I enter BIOS again, and set a static IP & netmask, no gateway, to IPMI to check out what it's saying, and reboot again. Time passes, and the IP never starts to reply to ping requests. Maybe the gateway was a required setting, even if it's not being used? Back into bios, and now it says:
BMC Firmware Revision: Unknown
IPMI STATUS: Not Working
Now I'm thinking maybe BIOS has to be updated in tandem with the BMC, but I didn't read anything about it, and the BMC didn't say anything about when i updated it via IPMI.
I cut power for all hardware changes. First removing all addon-cards, so I should be back in a working configuration, but no image via VGA. I remove all but 1 memory stick, same thing. Try another stick in another slot, same thing. So I add the 1080TI back again, and I get an image over HDMI.
I remove it again, clear CMOS, and switch to another battery while I'm at it. No changes.
I try booting the Ubuntu livecd over USB again. It loads GRUB, but hangs as soon as it starts to load the kernel, powering off the keyboard just as before.
Maybe this warrants a BIOS update, so I download latest BIOS_H12SSL-1B95_20220414_2.4_STDsp.zip and copy the contents to the EFI-partition of one of the linux drives I installed on earlier, and plug this in the USB->SATA adapter. Reading the help output for SUM.efi I realize it wants the IPMI password to flash a new BIOS, and thus as expected, it fails:
---8<---
FS0:\EFI\> flash.nsh BIOS_H12SSL-1B95_20220414_2.4_STDsp.bin ADMIN {{PASSWORD}}
[copyright header]
<<<ERROR>>>>
ExitCode = 101
Description = Driver input/output control failed
Program Error Code = 17.2
Error message: Send IPMI failed
---8<---
So, BMC seems broken, and I can't boot anything past UEFI-stuff anymore. Any Ideas?
No cratered IC as in the thread about "BMC Initiating", should I add pictures of the board?
I tried 2 other GPUs and another PSU too, just in case, no difference.