06:15:15 is there some docs for how omnios/apache is administrated now? It appears to use privileges (error EACCES [file_dac_write]) outside of any entry in httpd.conf 14:20:20 say I have a snapshot of a zone named rpool/zones/ldap0@zadm__pre_ldapmodify, and I would like to deploy a new zone from that snapshot named ldap1... what would be the most OmniOS way of doing that? 14:35:53 The official way would be to use 'zoneadm clone' instead of 'zoneadm install' 14:36:09 Configure as normal, and then something like 14:36:24 zoneadm -z ldap1 clone -s rpool/zones/ldap0@zadm__pre_ldapmodify ldap0 14:39:03 There are all sorts of caveats in documentation about the state of the source zone (you're supposed to halt it) 15:02:55 ptribble I had no idea that I'm supposed to halt it 15:09:01 I can't see that checked anywhere in code, though 15:09:53 And that might just be a documentation leftover from when zones were on UFS and you didn't have the atomicity of a zfs snapshot 15:15:41 Hm. I think zoneadm checks, but it's not clear why it's necessary 15:17:53 I never used the formal clone functionality; we had a hacky way of replicating zones from a snapshot via zfs send/recv instead, which avoids all the afs dependencies you have with clones, and works for building zones on remote machines as well as local 15:21:36 I wonder if zoneadm clone does send/recv or clone... 15:21:41 okay, time to play around 15:21:52 (I feel like the QA of illumos :P) 15:42:11 antranigv - that issue you raised about Hyper-V under Vultr is bizarre 15:42:23 andyf I just installed it again 15:42:29 andyf and I got the output 15:42:35 andyf sending to the issue now 15:42:38 Nice, let's find out what's going on! 15:43:25 and btw... I tried running it on my other machine... is there a reason for cpuid to print the output forever? :P 15:43:32 it's been going for 3 minutes xD 15:43:41 Yes, there are some bugs in cpu emulation 15:43:53 I did fix one bug recently in bhyve that would cause that 15:43:55 got it! it is running on bhyve... the other machine, that is 15:44:18 ah, probably illumos 16223 (fenix) 15:44:19 BUG 16223: bhyve returns bogus cpuid 8000_001D leaf (Closed) 15:44:19 ↳ https://www.illumos.org/issues/16223 | https://code.illumos.org/c/illumos-gate/+/3250 15:45:47 That's affected in FreeBSD as well, then. 15:45:55 Yes, the bug is there too 15:46:15 I have not sent it their way yet. 15:46:54 andyf here you go: https://github.com/omniosorg/illumos-omnios/issues/1432#issuecomment-1948671446 15:46:56 Ok, do you see the line in the output you shared for CPUID leaf 4000000? 15:46:57 CPUID 40000000:00 = 40000005 7263694d 666f736f 76482074 | ...@Microsoft Hv 15:47:08 Yup 15:47:11 The system is identifying itself as Hyper-V 15:47:17 that is so weird 15:47:35 I did confirm with support that it's KVM, and the BSD Loader says KVM as well 15:48:20 See https://github.com/illumos/illumos-gate/blob/master/usr/src/uts/intel/sys/x86_archext.h#L1780-L1788 15:48:32 For KVM, we'd see KVMKVMKVM or TCGTCGTCGTCG 15:48:45 Oh, BSD is showing KVM? Interesting. 15:49:17 If you have a conversation going with support, it would be worth sharing that output and asking them about the hypervisor signature! 15:51:40 andyf sir yes sir... I'll open a ticket again 15:52:21 In the illumos loader screenshot you added it does have QEMU and KVM in the smbios information. 15:54:01 This might be relevant https://community.netdata.cloud/t/vultr-linux-kvm-vps-but-netdata-detect-it-as-microsoft-any-idea/3535/3 15:56:11 Which makes it sound like they are running QEMU/KVM but hacking in some Hyper-V emulation. Cracking! 15:56:43 feels like browser user agents where everybody is just lying about everything 15:57:53 They identify as hyper-v just enough to make some windows drivers work, it appears, but not enough that we can actually boot with it. 15:58:04 It's enlightenment! 15:58:18 haha 15:58:23 I'm not sure how we could detect that.. 15:58:33 Well, maybe we can, depending on where it crashes. 15:59:22 Otherwise I'd suggest disabling the hv_vmbus driver as you're doing, or disabling platform detection entirely (which would be "echo set enable_platform_detection = 0 > /etc/system.d/platform") 16:45:45 andyf I guess that's because I'm using a custom OS. The FreeBSD ones say it's KVM indeed. But I do wonder, how is the BSD Loader detecting 16:46:02 andyf not an issue per say, but maybe we can document it somewhere on the website? 16:50:27 Does the illumos loader do the same thing as BSD? 16:50:39 I think it's getting it from the smbios information that the hypervisor presents 16:53:28 andyf apologies, when I say "the BSD Loader" I mean the FreeBSD loader in illumos :D 16:53:39 so the illumos loader, does indeed, say KVM 16:55:20 Ah, ok, sorry, with you talking about FreeBSD guests too I got confused. The illumos loader is a fork of the FreeBSD one. IF you still have a guest, you can use `smbios -t1` and `smbios -t2` to see the smbios information. That will say KVM/QEMU. 16:55:25 Vultr's platform is truly split-brain 16:56:14 andyf I can complain to the CTO, if needed :P 16:56:26 I took a snapshot of the VM so... checking again!!! 17:12:18 andyf here's some output: https://gist.github.com/antranigv/c015755a474dc70158e0747d2774d5b0 17:14:24 Yeah, Manufacturer: QEMU; Family: KVM 18:03:00 andyf so basically, some part of it is reporting KVM while the other Hyper-V? 18:03:47 Yes. It's KVM/QEMU but they have patched in a different hypervisor ID and emulated enough MSRs and other things to convince windows enlightened drivers to work. When we check during boot, it looks like Hyper-V. 18:19:58 andyf I see 18:24:04 Does anyone happen to have an omnios 048 vm with NVMe up? 18:24:11 (Or non-VM) 18:24:46 Yes, under propolis 18:25:44 Can you sanity check the ability to print the n_idctl field in the nvme_t gets data? ala *nvme_state::walk softstate | ::print nvme_t n_idctl[] 18:25:47 Context is https://illumos.topicbox.com/groups/developer/T5053860bd9418586-M78fe69364042867e3dac0573. 18:27:32 https://paste.ec/paste/EULM3Ml8#Sw1gO+YKvZ-cmhM7zk36IwtB8JB4BEgOWogTi4ZVVFl 18:27:45 OK, so it does owrk. 18:27:52 a lot of zeros, but the first few fields look good 18:27:53 I'll try to reply with the right syntax.