[06:45:04] <akosiaris>	 👋. FYI, we got our first kubernetes boxes using a GPT partitioning scheme. Those would be aux-k8s-worker200[6-9]. They are not yet put into production, but the imaging process was successful at the first try, so 🤞
[06:54:45] <XioNoX>	 akosiaris: scope creep, but are you planning on using the EFI version too?
[06:56:11] <XioNoX>	 of course I look and that's what you'r already doing! that's awesome
[07:01:39] <akosiaris>	 I do?
[07:01:47] <akosiaris>	 :P
[07:02:11] <akosiaris>	 jokes aside, yes, it's not just GPT, there's the EFI partition in there. 
[07:02:31] <akosiaris>	 The intent is indeed to boot them in EFI mode. How close we are to that, currently unknown
[07:11:07] <elukey>	 let us know if you need help!
[08:46:21] <akosiaris>	 well, it didn't work ofc. https://phabricator.wikimedia.org/T397415
[08:52:39] <elukey>	 akosiaris: that is probably my fault, supermicro has been really challenging.. in that case, I think that the option set for the NIC (to pxe boot) is not accepted by the redfish version
[08:52:53] <elukey>	 I'll double check later and report back, there may be yet another special use case
[08:53:29] <elukey>	 for example, in some cases if you use "legacy", in the BIOS you need to set "legacy" for PXE, in other cases "pxe"
[08:53:37] <elukey>	 and for EFI I imagine the same sigh
[08:53:48] <elukey>	 is it documented? 
[08:54:05] * elukey doesn't answer 
[08:55:58] <elukey>	 like "BIOS: P2_AIOMAOC_A25G_b2SLAN1OPROM is set to 4, while we want EFI"
[08:56:44] <elukey>	 this is a first though, usually setting "EFI" gets rendered to "4" but it works, in this case it doesn't
[08:57:14] <elukey>	 and this is Config C, should work
[08:57:17] <elukey>	 ufff
[09:30:11] <elukey>	 akosiaris: fixed the issue, I am completing the cookbook so you are unblocked afterwards
[09:30:22] <elukey>	 ETA 5 mins
[09:31:36] <akosiaris>	 wow, that was fast, I didn't expect someone would dive into this
[09:32:04] <elukey>	 it is sadly something that I have been working for the past year, it still bites me
[09:37:11] <elukey>	 akosiaris: go ahead!
[09:37:31] * akosiaris trying again
[09:37:32] <elukey>	 the patch that fixed the cookbook is not yet merged, so if you need to run it on other nodes we'll have to use test-cookbook
[09:37:46] <elukey>	 akosiaris: you can skip the cookbook now, already executed
[09:37:53] <elukey>	 basically it is now set for UEFI etc..
[09:37:59] <akosiaris>	 ?
[09:38:05] <akosiaris>	 oh I thought you fixed the cookbook
[09:38:09] <akosiaris>	 ok, cancelling
[09:38:16] <elukey>	 yeah I had to use test-cookbook
[09:38:24] <elukey>	 and executed it etc..
[09:38:31] <akosiaris>	 I have like another 3 hosts to test it on later on
[09:38:32] <elukey>	 the ETA was for the completion of it :)
[09:38:44] <akosiaris>	 so I can be your guinea pig I guess
[09:38:48] <elukey>	 okok, it should be merged by then
[09:44:55] <akosiaris>	 ok I just rebooted it via the SMASH-CLP shell
[09:44:57] <akosiaris>	 waiting now
[09:55:18] <elukey>	 wait a sec, shouldn't you reimage?
[09:56:17] <akosiaris>	 should I? What else is there that the reimage would do? The GPT partitioning scheme is there so is the efi partition
[09:56:27] <akosiaris>	 I mean I can, but why?
[09:56:56] <akosiaris>	 also how on earth do I get a console on supermicro...
[09:57:30] <akosiaris>	 cd sol1 ; start
[09:57:33] <elukey>	 https://wikitech.wikimedia.org/wiki/SRE/Dc-operations/Platform-specific_documentation/Supermicro_Documentation
[09:57:38] <akosiaris>	 this SMASH-CLP thing is weird
[09:57:50] <akosiaris>	 I remember it from like 15 years ago, it was weird then, it is weird now
[09:58:14] <elukey>	 wait a sec, I am confused - I thought you needed to reimage to apply the new partition scheme etc..
[09:58:42] <akosiaris>	 partitioning scheme was applied straight from the first image
[09:58:42] <elukey>	 so the host was reimaged the first time with the right EFI partman reciped, but not provisioned with UEFI?
[09:58:53] <akosiaris>	 yes. At least that's what I think right now
[09:59:00] <akosiaris>	 could be wrong
[09:59:20] <akosiaris>	 it's starting up, stuck currently at >>Start HTTP Boot over IPv4.
[09:59:23] <akosiaris>	 lol?
[10:00:28] <akosiaris>	 ok, got in a UEFI interactive shell
[10:00:35] <akosiaris>	 it clearly did not like something 
[10:00:53] <elukey>	 so with supermicro and UEFI we have an open issue related to the boot order, the TL;DR is that the first time one needs two reimage to get it to a successful state
[10:01:01] <akosiaris>	 elukey: I 'll reimage as you point out, it's probably more than just the GPT + EFI partition thing
[10:01:16] <akosiaris>	 it probably needs to set the efi vars for the boot manager anyway
[10:01:30] <elukey>	 I think so yes.. if it happens that the puppet run step breaks it may be due to the aforementioned issue, and another reimage will be needed
[10:01:39] <akosiaris>	 ok, let's try that out
[10:01:40] <elukey>	 it happens only the first time that we reimage
[10:02:00] <elukey>	 and supermicro is still not recognizing the problem, despite a lot of attempts from us 
[11:19:38] <akosiaris>	 akosiaris@aux-k8s-worker2006:~$ sudo dmesg | grep efi
[11:19:38] <akosiaris>	 [    0.000000] efi: EFI v2.80 by American Megatrends
[11:19:43] <akosiaris>	 elukey: habemus efium!
[11:27:02] <akosiaris>	 btw, ofc you were correct. Proper GPT partitioning with an EFI partition does not mean that the efi partition gets grubx64.efi installed. That's the job of the installer and the installer needs to be started in uefi mode to do that. 
[12:46:42] <elukey>	 \o/
[12:47:00] <elukey>	 thanks a lot for testing UEFI