[00:13] no, there is no 'default - always right', also the bits differ for files vs. directories [00:14] I interpreted his question to be about / specifically [00:14] (it is totally possible that 755 on files is what you want, though) [00:14] oh, right, i didnt [00:15] iif it's just about the uppermost directory on any ubuntu system (which also happens to serve web content) then i agree [00:16] tomreyn: I don [00:16] whups [00:16] oh he left [00:16] if he ran chmod -R 777 / then the eeasiest way out is going o be renting a new server and moving over the files you care about :) [00:16] I don't know if you saw the previous msg, but just to loop back [00:16] I had to set fb=false [00:17] nofb doesn't work for whatever reason even tho it's an official kernel parameter [00:17] drab: alrighty, thanks, and glad you found out. [00:21] any tips on must do tweaks for qemu performances? [00:22] it might be useful to pass in the 'current cpu' as the cpu type; that way all the instructions you -do- have can be used [00:22] I think the defaults are conservative so you can migrate from one compute host to another [00:30] k, thanks [00:30] two things I've seen is the use of virtio in disk devices and network devs [00:30] but still not clear how those work [00:31] instead of emulating ten-year-old hardware [00:31] they just shove bytes around :) [00:31] :) [00:32] i.e. use whenever you can [00:32] well that's part of the thing, how do I know if I can? I don't get what's special about them that makes it possible or not to use them [00:33] you need guest support [00:33] ah, ok, that's one bit I missed [00:33] linux has it, but windows doesn't until you install drivers [00:33] most non-terrible OSes have support for virtio of differentl flavors [00:33] oh, that's np, I'm just doing linux [00:33] even, just ubuntu xenial (for now at least) [00:33] so nothing special to install there? it just works [tm]? [00:34] it should [tm] [00:34] heh :) [00:35] what about virtio-scsi ? I've seen that floating around on a few pages as "better", but in various examples I only saw virtio-blk-pci [00:38] hmm i'm not sure there, maybe sarnold knows which one to prefer. my guess is that scsi is more overhead but more compatible. like a (PCI bus attached) nvme conmpared to a (SCSI controller attached) ssd [00:38] I hadn't heard of virtio-scsi before [00:39] it popped up on this guide: https://doc.opensuse.org/documentation/leap/virtualization/html/book.virt/cha.qemu.host.html#kvm.virtio-scsi [00:39] in the recommended host settings [00:39] that guide is actually the most complete I've found [00:39] suse invests a -lot- in great docs [00:40] aha! scsi passthrough [00:40] that doesn't try to say "oh you wanna run qemu plain? let me tell you about libvirt" [00:40] if you -want- scsi passthrough that'll be better indeed. but that's a specific thing. [00:40] :? [00:40] gotcha [00:41] tbh I'm kind of confused by scsi... back in the days you had to have a scsi card and scsi devices... now a usb key seems to go thorugh the scsi layer [01:22] drab, scsi is just a common verb set to move blocks. no sense in reinventing the wheel, again. [01:22] sarnold, I guess virtio-scsi is just more better: https://www.ovirt.org/develop/release-management/features/storage/virtio-scsi/ [01:25] ppetraki: nice page, thanks [01:26] ppetraki: I've never heard of the ~30 device limit of virtio-blk before [01:26] that would certainly limit some operations :0 [01:27] sarnold, I guess it's a PCI limit of whatever bridge they virtually implemented [01:28] sarnold, now you can do multpath in a vm! [01:29] ppetraki: fair enough, thanks [01:29] * sarnold fails over [01:29] I'm trying to find out why it doesn't let me use virtio-blk, or if it's just a naming problem [01:30] ie when I say "virtio" in the device definition it's really using virtio-blk-pci [01:30] it takes virtio-net-pci for the network card, but not virtio-blk for the img file [01:31] drab, how was the image file built? raw image? qcow2? [01:34] drab, so my favorite thing to do with qemu and the options is to cheat. I configure something using virt-manager the way I think it should look, start it, then look at the qemu args passed to it. [01:35] ppetraki: raw image on top of zfs (hence avoiding qcow2, as I understand that'd be twice copy on write and not worth it) [01:35] drab, ok, good, did you tell qemu format=raw? [01:36] ppetraki: I'm having problems cheating, tried to get libvirt going on my desktops and just got all kinds of fails [01:36] I did [01:36] erg! [01:36] I don't want use libvirt long term so I didn't bother trying to figure out what I was doing wrong [01:37] I don't want to use libvirt long term either, neither did openstack, but dammit it just wont die... [01:40] lol [01:40] drab, ok I cheated for you http://pastebin.ubuntu.com/24430578/ [01:41] drab, now you can shove this in a script and parameterize what you need [01:42] drab, I just made it install from whatever iso I could find, create a 4G RAM 2 CPU system with a 20G root [01:43] holy cow, looks more like bad time reading than a command line [01:43] thanks, much appreciated [01:43] drab, this looks like the line that matters the most "-drive file=/var/lib/libvirt/images/centos7.0.qcow2,format=qcow2,if=none,id=drive-virtio-disk0 -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x7,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=2" [01:43] oh I don't read anymore, my eyes hate me [01:43] lol, bed time reading, Freudian slip I guess [01:43] I hear you [01:44] today I got lucky, I got to spent 4hrs in a basement pulling fiber <3 [01:44] spend* [01:44] so the -drive switch creates the vlun and also instantites a label "drive-virtio-disk0" and then attach that to the drive we want in the -drive switch [01:44] I mean -device switch [01:45] yeah, see, you have virtio-blk-pci there, in the device definition [01:45] I would make the boot index lower because this thing is "booting from CD" and change the format to raw, other than that "its simple" [01:45] somehow that doesn't work on my test machine for some reason I don't get [01:45] are you root? [01:45] yeah, looks ok, also looks like it's got the stuff I was wondering about re monitoring socket, so will check that out too [01:46] drab, sudo kvm-ok for me [01:46] lemme try again with stuff from the example, the onliner I had was clobbering device and drive, maybe shorthand or hold syntax [01:47] yeah kvm is ok [01:47] drab, ok. sounds like a plane [01:47] I mean plan [01:47] * ppetraki is tired [01:59] ppetraki: it kind of worked, I had to append an if=none to the drive which I don't get, but otherwise it went through [01:59] qemu-system-x86_64: -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x7,drive=d-v-disk0,id=dddisk0: Drive 'd-v-disk0' is already in use because it has been automatically connected to another device (did you need 'if=none' in the drive options?) [02:00] that's the error I got [02:02] drab, its open sore, it's not supposed to make sense [02:02] drab, that sharp edge is supposed to be there ;) [02:05] open sore, that's gonna stay with me for a while :D [02:05] drab, I got it from engineers way older and saltier than I [02:13] o/ [02:15] .o/ [02:16] thanks again man, much appreciate the help, have a good rest of the day [06:41] Good morning === JanC is now known as Guest51396 === JanC_ is now known as JanC === Ale is now known as Guest33116 [12:48] hi [14:50] trying to pxe install to a nvme, someone in #u said try here. onboard nvme ssd thing. I somehow got xenial to boot last year. wiped it trying to install zesty. installer runs, I get these partitioning options: [14:51] the installer gave me 1 big partition. text mode install screens about partitioning https://veyepar.nextdayvideo.com/static/temp/d.html [14:52] the installer does install files to it, but when I boot: "insert boot media" https://veyepar.nextdayvideo.com/static/temp/d6-0.png [15:12] CarlFK: just to confitm, you said yuo had it working? [15:12] becuase I had that same rpoblem and it turned out the mobo/bios just didn't support it [15:12] (this was a SM server and I chcked with them/bios engineers) [15:12] drab: yes working. no idea how. i play with a bunch of machines. [15:13] but it otherwise exibhited the same behavior as you describe, reason being the boot is done with kernel from pxe, not from nvme, and by the time that's loaded /root can be loaded from nvme [15:13] CarlFK: 100% sure you weren't booting from something else? a usb key, a cd, a small /boot on another disk... etc? [15:13] don't mean to be insisting, but it's kind of crucial :) [15:14] and we'd be all wasting time if it wasn't actually possible and there was something else lost in translation [15:14] possible -> straight boot, indirect is obvious possible [15:15] "no idea how" is the part that kind of worries me :) [15:15] have you checked the mobo's manual, does it say it can boot from nvme? same for bios [15:15] what mobo is it? [15:15] and what bios version [15:16] also what nvme device? is it plugged straight into the mobo or are you using some kind of pcie adapter? or is it a pcie nvme device to begin with? [15:25] /sys/devices/virtual/dmi/id # cat board_name [15:25] H170M-D3H-CF [15:26] http://www.gigabyte.us/Motherboard/GA-H170M-D3H-GSM-rev-10#support-manual [15:26] closest I could find, doesn't matter, pdf wont' dl [15:26] nor the bios [15:27] nvme plugged into slot on mv. samsung something. 512gig [15:28] mb lists it as a boot option. I suspect I had the efi stuff setup before, and it listed both efi: and non as 2 boot options [15:28] ok, that was my next thing, nvme support generally is uefi only, not bios. are you booting in bios or uefi mode? and was linux installed in eufi mode? [15:28] I am starting to think the u installer di will do efi if it sees the fat partition. so maybe I have to create that first? [15:29] CarlFK: ime it does efi if the bios is setup for efi only/the installation is started as efi [15:29] if you're pxe'botting are you passing efi files to it? [15:29] booting* [15:29] oh.. pxe.. no. I bet thats what I need to find [15:30] yeah pxe needs additional steps/work to boot in efi mode and do an efi installation [15:31] if you're running vanilla pxe you're likely installing in bios mode [15:31] you could try to install with a CD if you can, might be quicker to verify [15:31] and make sure your bios is set to efi or that when the prompt comes up you select to boot the CD as efi [15:32] which will get you an efi install [15:32] at least that's afaicr, my efi foo id generally weak, I'm old and so is my hw :) [15:33] anybody knows a workaround for this bug: https://bugs.launchpad.net/ubuntu/+source/memtest86+/+bug/1564740 [15:33] Launchpad bug 1564740 in memtest86+ (Ubuntu) "Booting kernel failed: Invalid argument" [High,Confirmed] [15:33] what I don't get is why memtest works fine on both vbox, a bunch of desktops, but not the SM machine [15:33] some of the desktops are also intel [15:33] they are all in bios mode [15:34] why would you run memtest in vbox [15:34] pxe booting the memtest image served via tftp [15:34] it's virtual memory [15:34] but why would you do that in a vm [15:34] ikonia: to test that the pxe infra is working and passing up the right files [15:34] so I test that first, thant est on a couple hw machines/desktops, then test on the server [15:35] what happens when you try to boot it for tftpboot [15:35] that why I cover all classess of hw and scenarios, since pxe must work on virtual too to provision containers [15:35] except I don' [15:35] I'm not actually testing the ram [15:35] except I don't have a cd drive.. but leme find pxe efi stuff... [15:35] ikonia: it boots fine on the desktops and the VM, it just fails on the SM server [15:35] like in that bug [15:35] it complains about an invalid argument [15:35] SM server ? [15:35] supermicro server [15:35] what's an SM server [15:36] ok - so thats most likley the pxe setup on the card [15:36] the invalid argument is known on some network cards [15:36] you can get around it with a setting on the tftpserver [15:36] uhm, ok, interesting, hadn't heard of it [15:36] I can't remember which one it is off the top of my head [15:36] great, ok, that's still a useful hint, thank you [15:37] I can't remember which way around it is, either some cards don't fully comply and expect an argument that isn't needed, or they are over strict and demand an argument that isn't needed [15:37] I can't remember which way around it is [15:44] CarlFK: the bios and manual download work fine from my location. should i store a copy for you somewhere? [15:45] (and have you tried it from a different location, yet) [15:45] tomreyn: I tired a few of the area links. wget says 2017-04-22 10:45:02 (1.47 MB/s) - Read error at byte 31119/13554428 (Connection reset by peer). Retrying. [15:46] oh hey, 3rd times a charm. got it. [15:52] http://archive.ubuntu.com/ubuntu/dists/zesty/main/uefi/ thats all I see around install stuff [15:59] ikonia: ok, found a workaround [15:59] I don't know why the other wasn't working and what was up with what you mentioned, couldn't google anything about it [16:00] but I changed and instead of trying to boot the memtest binary [16:00] I used memdisk to load the memtest.iso from passmark [16:00] and that works across all environments/hw classes so good enough [16:00] takes a few extra secs to get the iso compared to just the bin, but whatever [16:01] CarlFK: if you don't need to pass preseeding, you could do what I just did [16:01] and load the entire iso from network [16:01] that'd be the same as booting from CD [16:01] vs booting a kernel and going straight into install [16:02] I am a little confused. not even sure what terms to use any more. I either boot bios or efi, right? [16:02] right [16:02] well, what do you use? [16:02] bios or efi? [16:02] but there's two pieces, both the bootable media and the bios must be on efi/or bios [16:03] you can have an efi that fallsback to bios [16:03] and stuff like that [16:03] so the boot part needs to be configured on efi, boot the install media as efi, and then the install media needs to install efi compatible boot loader [16:04] mostly this magically happens, ie ubuntu install CD will pick up a boot from efi and isntall efi, at least that's my experience on desktops, but with pxe boot is a little more complicated [16:04] and maybe all has to be signed too [16:04] since you decice what to serve [16:04] that too, le sigh [16:04] brb [16:05] patdk-lap: #ipxe says NiXZe: the rom usually has both efi and pcbios, however the configuration to enable them is confusing to say the least in most bios/firmware implementations [16:05] so I am trying to use efi, but I may not be. [16:05] heh? it's supersimple in every one I have seen [16:06] you just go into the settings, and select uefi boot, and disable legacy boot [16:06] now your in uefi/efi mode [16:06] have you looked at the firmware? ;) [16:06] enable legacy and you are back to old fashon bios boot [16:06] no, cause I have no idea what device you have [16:07] #ipxe (05:50:16 PM) Shiz: and sometimes the option is there but the efi rom simply is not [16:07] I pick efi options, but seems I am ending up in legacy [16:08] is there any way to tell from the installer busybox shell? [16:08] what system is this? [16:09] /sys/devices/virtual/dmi/id # cat board_name --> H170M-D3H-CF [16:09] http://www.gigabyte.us/Motherboard/GA-H170M-D3H-GSM-rev-10#support-manual [16:09] closest model I could find [16:21] oh wait [16:21] you want pxe boot? [16:21] yes [16:21] it is *rare* for nic pxe boot to work in uefi mode [16:21] I have seen some the last month [16:21] but before that, I would have said it wouldn't work at all [16:21] can I chain load ipxe? (which I guess has efi support) [16:21] isn't pxe that thing for 32 bit systems that lets them access larger addresses [16:22] jeremy_carroll, heh? [16:22] a30 enable? [16:22] pxe = load kernen/init from tftp server [16:22] maybe i'm thinking of another acronym then [16:22] nic rom has a little tcpip stack [16:22] yes, you could boot ipxe, if ipxe has a uefi mode [16:22] but ya, not the nic rom [16:23] ya, that board says nic does not support uefi [16:23] storage boot option set to uefi only [16:24] csm support set to disabled [16:24] (csm support is the, if uefi failes, fall back to legacy bios boot) [16:25] btw - I have the box the board came in: ga-h170m-d3h - no mention of cf [16:26] if I am currently able to pxe boot the installer, do I need to change anything ? [16:27] er [16:27] to accomplish what? my znc buffer doesn't go far enough back [16:27] if you want to use uefi? you must boot uefi [16:27] or the installer will not setup uefi install [16:27] (11:22:46 AM) patdk-lap: yes, you could boot ipxe, if ipxe has a uefi mode [16:27] if you don't care, then don't worry, whatever booted it will install and make work [16:28] if you want to make absolutely sure you're booting in uefi mode, check for the presence of /sys/firmware/efi/efivars [16:28] I have never uefi booted ipxe before, as the systems I use it on don't have uefi support, only use old systems with it [16:29] can I boot from the nvme? (ssd on the board) [16:29] er, in legacy mode [16:30] ask your bios [16:30] I couldn't figure that out from the manual. [16:30] or the bios [16:31] currently when it boots the ssd i get "no boot media" [16:31] I have booted all my nvme systems without uefi and with uefi both [16:31] i can pxe boot the installer into rescue mode, mount it. grub-installer it.. no error, but still not boot [16:58] zesty installer BusyBox v1.22.1 (Ubuntu 1:1.22.0-19ubuntu2) ~ # ls /sys/firmware/ -> acpi dmi memmap [17:01] same for /target/sys/firmware/ [17:16] so I pxe booted, I am in legacy mode, both the installer shell and what the installer installed. bios says it will boot the ssd in legacy mode. what do I do to install grub? [17:16] (pretty sure there is a bug here, but I dont' know what I am doing well enough to just file a issue ) [19:26] https://en.wikipedia.org/wiki/EFI_system_partition "some UEFI implementations immediately switch to the BIOS-based CSM booting..." [19:27] who knows what mode we started in? [21:29] re [21:29] CarlFK: did you manager to get it to work? [21:29] manage* === haasn is now known as hanna