Thank you to everyone who has responded. Every response had helped move this issue forward as well as teach me something new. I want to send an update on progress to everyone as well as provide feedback to help future user with similar issues.
The link sent by YANG Song was very helpful. I am now testing on 8 different nodes, which helps ensure I am not dealing with a single piece of bad hardware. A new update in the story background is that the CentOS install is no longer creating a /boot/efi partition on the hard disk. I'm not sure why this is the case but I suspect it is a change made by a colleague. (I will try to send a follow-up if I find out the cause.) I verified BootModes.SystemBootMode is UEFI (this appears to be the default, I didn't have to set it on any nodes.) I added "Legacy Only" to the BootOrder.BootOrder. After some playing around with different ordering, I am leaving the BootOrder as: BootOrder.BootOrder=Legacy Only=Hard Disk 0=PXE Network As Jarrod mentioned, booting from hard disk first avoids all the PXE network checks that occur at boot. To re-install a node, I use rsetboot to force a PXE attempt at next boot. I should reiterate that the OS is no longer installing the EFI boot partition. This is very important because setting "Legacy Only" in the BootOrder will cause a machine to not boot IF said machine has an EFI boot partition. At this point, I still have several questions unanswered, but am putting these off till I have more time to research and test more as well as learn more about EFI. Unanswered questions are: How to make CentOS install (or not) a /boot/efi partition during OS install? When CentOS installs a /boot/efi partition and adjusts the boot order, why does the new boot option (named CentOS) fail to boot? Removing the "CentOS" boot option manually allows the machine to boot into the freshly installed OS. Thanks again to everyone! On Tue, Aug 2, 2016 at 8:07 AM, Jarrod Johnson <[email protected]> wrote: > For what it is worth, I'd say that so long as you have working 'rsetboot', > I actually like the 'boot from hard disk first on normal boot' behavior. > It's a touch faster and rsetboot can override. > > > > That's my personal opinion on that mode. > > > > *From:* Andrew Loftus [mailto:[email protected]] > *Sent:* Friday, July 29, 2016 4:56 PM > > *To:* xCAT Users Mailing list > *Subject:* Re: [xcat-user] Boot order changed after CentOS install on > Lenovo 3550 > > > > # pasu backup01 show BootOrder > > backup01: BootOrder.BootOrder=CentOS=PXE Network=Hard Disk 0 > > backup01: BootOrder.WolBootOrder=PXE Network=CD/DVD Rom=Hard Disk 0 > > > > What's interesting is that the UEFI update is expected behavior, which > seems to conflict with the preferred behavior for an xCAT deployed machine > is to always boot from PXE first, to allow a re-install when necessary. > Otherwise, an extra step is required to force a new installation, namely, > rsetboot. > > > > Am I correct in my thinking? > > > > I'm going to try Arif's suggestion of setting > > BootModes.SystemBootMode=UEFI Mode > > and see how that goes. > > > > On Fri, Jul 29, 2016 at 3:12 PM, Jarrod Johnson <[email protected]> > wrote: > > > > Can you do pasu <node> show BootOrder > > > > You could, if you'd like, change the boot order to 'Hard disk 0' with > asu/pasu and it should work (centos 7 will work that way, centos 6 will > not). > > > > I may also like to see asu show all if you can't find hard disk at all. > > *From:* Andrew Loftus [mailto:[email protected]] > *Sent:* Friday, July 29, 2016 4:07 PM > *To:* xCAT Users Mailing list > *Subject:* Re: [xcat-user] Boot order changed after CentOS install on > Lenovo 3550 > > > > Thanks for the information Jarrod! > > > > There is no special partitioning and only one "visible" hard drive (there > are two physical drives setup in a hardware mirror). > > > > I'm relatively new to UEFI systems. Perhaps if I understand it better I > can troubleshoot it better. > > Where can I learn more about the UEFI install procedure that you mentioned > above? > > > > > > > > On Fri, Jul 29, 2016 at 2:40 PM, Jarrod Johnson <[email protected]> > wrote: > > Just using the normal storage local? No SAN attached storage or anything? > > > > For reference, on install UEFI compatible systems look at the partition id > they write their boot loader to. They take the partition uuid and tell > UEFI firmware 'hey, next time boot to this file on the partition with this > UUID'. > > > > If it for some reason selected a storage device that is *not* visible from > UEFI, this behavior would be seen (It's looking for a UUID that doesn't > exist). > > > > Is there a custom partition plan? Are there multiple disks? Some > versions of CentOS will struggle if '/boot' and '/boot/efi' get split up, > for example. > > > > *From:* Andrew Loftus [mailto:[email protected]] > *Sent:* Friday, July 29, 2016 3:36 PM > *To:* [email protected] > *Subject:* [xcat-user] Boot order changed after CentOS install on Lenovo > 3550 > > > > xCAT version: 2.11.1 > > OS: CentOS Linux release 7.2.1511 (Core) > > Hardware: Lenovo X-series 3550 > > Install type: diskful install > > > > Node PXE boots successfully and OS install completes successfully. Upon > reboot, the machine fails to boot because it can't find a valid OS. Using > rcons to get to machine console, we find a new boot option named 'CentOS' > and it is first in the boot priority list. > > > > Does anyone know where this comes from? > > Why it's there? > > How to fix it or prevent it from getting set in the first place? > > > > Even a gentle shove in the right direction to troubleshoot this would be > greatly appreciated. > > > > Cheers, > > --Andy > > > > >
------------------------------------------------------------------------------
_______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user
