From bug 983302: It looks like CentOS 6.5 does not work out of the box on either our old IX700C's, nor on the new iX21X4's (multinodes). On the latter, I get "ERST: failed to get error log address range" which everything I've read suggests requires a BIOS update. Disabling ACPI in the bios gets a blank screen with a blinking cursor. Setting ACPI level 1.0 in the bios gets the same error message. With erst_disabled in the kernel command line: GHES: Poll interval is 0 for generic hardware error source: 1, disabled. with acpi=off as suggested in http://www.supermicro.com/support/faqs/faq.cfm?faq=15594, it stops booting after loading device-manager (with no error message)

I submitted an ixsystems support request. The BIOS on these hosts is five years old, so I'm almost certain the answer is "update". So, I'm on the hunt for a DOS image..

And of course, dosemu doesn't work. I'm setting this aside for the moment.

For my reference: http://derek.chezmarcotte.ca/wp-content/uploads/2012/01/FreeDOS-1.1-USB-Boot.img.bz2 mounted with kpartx -v -a FreeDOS-1.1-USB-Boot.img mount /dev/mapper/loop0p1 /mnt/usb (cd /mnt/usb; unzip /tmp/X8SIT2_627.zip) umount /mnt/usb However, the result is a USB image with a partition table, which is not something that iKVM knows how to handle.

On the production hosts: [root@b-linux64-ix-0001.build.releng.scl3.mozilla.com ~]# facter | grep bios bios_release_date => 05/27/2010 bios_vendor => American Megatrends Inc. bios_version => 1.1 On the relabs hosts, the BIOS version is v02.67, which is what shipped on the device, and copyright 2009.

b-linux64-ix-0001 is an ix700C, duh.

This one is an ix21x4: [root@talos-linux64-ix-036.test.releng.scl3.mozilla.com ~]# facter | grep bios bios_release_date => 06/27/2012 bios_vendor => American Megatrends Inc. bios_version => 1.2a This one seems to hang at "detecting hardware..." during the Anaconda start-up.

Entering the BIOS at startup confirms the BIOS version and date uncovered by facter. After the Anaconda failure at startup, the existing Ubuntu boot no longer works, though. On normal (non-PXE) boot: error: file not found grub rescue>

Well, it seems that the kickstart *was* working, just running on the external graphics card. I hates the external graphics cardss. So that time around, it had gotten as far as erasing the disks when I rebooted. The second time around, I let it run for a while and it eventually bombed out on a missing PUPPET_PASS, but did install 6.5. On a reboot: ERST: Failed to get Error Log Address Range so, that's fun. It looks about the same, in other words.

I forgot to mention in comment 0, without 'quiet' in the kernel command line, I also see ACPI Exception: AE_NOT_FOUND, No or invalid critical threshold (20090903/thermal-386) just before the failure. This seems to be harmless - https://access.redhat.com/site/solutions/726383 RH blames the ERST error on firmware, too: https://access.redhat.com/site/solutions/170253

(In reply to Dustin J. Mitchell [:dustin] from comment #6) > [root@talos-linux64-ix-036.test.releng.scl3.mozilla.com ~]# facter | grep > bios > bios_release_date => 06/27/2012 > bios_vendor => American Megatrends Inc. > bios_version => 1.2a This is the latest BIOS. I'm not sure how I missed that earlier. So BIOS upgrade is not a fix.

> On the relabs hosts, the BIOS version is v02.67, which is what shipped on > the device, and copyright 2009. I was wrong here, too. That's the version of the SETUP utility. The BIOS itself is 1.2, build date 12/22/11. That's arranged in the display to make it look like it's the motherboard version and build date. So, not the latest, but it was probably the latest available when these hosts were purchased in September 2012. And something's not making sense here. I just re-imaged ix-mn-2 with CentOS 6.5, and it took forever and an age to do the install, appearing to "hang" during install just like comment 8. I let that run for a while and it eventually did restart, and is currently displaying the "ERST" message. However, it got a certificate and ran puppet and I can login, despite that message staying onscreen and no getty "login:" prompt. I need to compare this to some other ix multinodes -- my impression was that these hosts looked the same as any other, which is to say, had a login prompt onscreen. My impression was also that these hosts displayed the usual anaconda mess while kickstarting. Perhaps both of these impressions were incorrect, in which case this is all a waste of time.

My impressions were correct -- CentOS 6.2 displays everything on the onboard video, while 6.5 switches to the add-on video, leaving only a confusing not-really-an-error message onscreen. So this works fine. Now that I'm on a roll, I'll take a whack at the ix700C's too.

Well, at best that will involve building a new kernel/initrd for anaconda, which doesn't sound worth it for ancient hardware.