=== Isla_de_Muerte is now known as NwS [01:53] cpaelzer: yeah i was going to use it to check whether upstreamahd fixed the broken qemu-img convert -O vmdk wrt esx5 [01:54] cpaelzer: I do remember it was a pain to keep running at times, so I'm not complaining if it's not worth your time. [01:54] Although, it did also save time at merges [01:55] cpaelzer: how were you doing it? were you keeping the packaging for the daily in its own git tree? [06:33] hallyn: yeah I started an own git tree (owned by the repos user) which was a stripped down debian/ dir [06:33] hallyn: and then set up a LP recipe which pulled together repositories as needed [06:33] hallyn: the recipes are also in the git of the LP user [06:34] hallyn: I didn't delete anything, just disabled the recipes and added the statement why I disabled it on the ppa description [06:35] btw - good morning everybody [07:11] Good morning [11:21] /w/win 10 [11:21] oops, sorry [13:03] I have a question... what is the benefit of using netplan in newer ubuntu-server releases? [13:08] from what I've read it does not work with openvswitch yet.. even so.. is yaml config so appealing these days? === jelly-home is now known as jelly [14:01] Prototip: it's not a matter of whether it's YAML. The idea is to have a single common place to write networking configuration so users don't need to learn systemd syntax and NetworkManager syntax and something else, etc. [14:02] Prototip: so you can configure your networking the same way you did for ifupdown (sure, the syntax and grammar are a bit different, but we hope intuitive enough), when ifupdown started to show issues with complex network setups with bridges, vlans, etc. involved [14:06] wait, what did I miss, /etc/network/interfaces is not what it used to be ? [14:07] because editing that is how I did/do all my server interface config [14:11] Ussat, in zesty and up, ifupdown is no longer installed by default. Instead installers write out netplan yaml configuration file, which then on server by default uses systemd-networkd to bring networking up. [14:11] ... [14:11] Ussat, on upgrades, as usual, existing configuration and installed packages are preserved. [14:12] well, good thing mt 16.04 LTS isnt changing anytime soon [14:12] Ussat, ifupdown is racy and broken with bonded vlans among many other things =/ it really is not reliable at all. [14:12] Ussat, 16.04 LTS ships netplan as a supported, but not default, option. [14:12] right, reding now [14:13] I assume next LTS it will be the default [14:13] cyphermox: if networkmanager would know how to parse /etc/network/interfaces, the problem would be solved already. [14:13] Ussat, it is default in zesty and bionic, yes. [14:13] Prototip: no, because it did and that was a load of pain. [14:14] Prototip, no. as /etc/network/interfaces allows arbitrary code execution / it is turing complete scripting language. network-manager xml / networkd units are not. [14:15] xnox: oh, you mean pre-up and post-up scripts.. yes, that can get ugly.. [14:16] xnox, cyphermox: I hope that by 18.04 netplan supports openvswitch too. [14:17] 16.04 is still latest LTS right ? [14:17] I am all for switchover to better solutions as long as they do not downgrade functionalities [14:17] Prototip, also arbitrary hooks, the accept arbitrary any-case additional keys, with silent substitution of _ and - [14:18] Ussat, 16.04 LTS is the current LTS, we are currently developing bionic which will become 18.04 LTS in April - 2018-04 [14:18] 16.04 is supported for 5 years, plus to be announced paid-only Extended Security Maintainance (ESM). [14:19] given how popular ESM is for trusty 14.04, I expect there will be 16.04 LTS ESM too [14:19] Well, guess better grab a 17.X server and play with it [14:23] I installed a 14.04 server for someone, and cant wait to upgrade it [14:24] compdoc, well 16.04 has been out for a long time =) to upgrade 14.04 machines.... [14:24] compdoc, unless you are waiting for 18.04 to release to upgrade 14.04 -> 16.04 reboot -> 18.04 reboot in one go. [14:26] I cant justify the cost. I have to wait for it to expire [14:28] screw that, just clean install a8.04 at that point [14:28] 18 [14:28] bah [14:28] need more coffee [14:40] `apt upgrade` is stuck on the dpkg unpack gitlab-ce step. htop shows the status as D (uninterruptible sleep, usually disk I/O) and the disk usage of the process is minimak [14:41] It writes a few hundred MB/s for some amount of time and then just goes defunct like this [14:41] Nothing has changed since then [14:41] No log, no dmesg output, nothing [14:41] Any idea what could be going on and how to fix it? It's rather annoying having `apt upgrade` hang forever [14:49] Heh. just completed [14:49] Just spent forever doing literally fuckall [15:11] cpaelzer: ok, i *may* see about making that live again. though i guess now that qemu can mainly merge from debian it's far less useful. [15:16] hallyn: yep [15:16] hallyn: I won't stop you maintaining the ppa if you want, it just lost its value for me [15:17] hallyn: it worked when I was merging libvirt/qemu anyway as I knew what changes were currently in flight [15:17] hallyn: but then the usability falls of and maintenance rises [15:17] hallyn: on the last cycle it didn't give me huge plus on the "easen the next merge" thing [15:18] hallyn: and it had just 1 download. :) [15:30] dpb1: but that one download may have been inordinately helpful :) [15:31] yeah it would probably make more sense if anything to collaborate with debian on a single bleeding edge package; and really that's not necessary [15:31] mjt does too good of a job :) [15:33] hallyn: hah === JanC_ is now known as JanC [17:32] coreycb: I should have all of the oslo* in PPA before I eod (apart from messaging - see note) [17:35] Do anyone have a good script to backup volume groups [17:36] vgscan, foreach create snapshot, for each snapshot dd if=snapshot of=/backup/snapshot.img then remove snapshot? [18:01] jamespage: ok sounds good! [18:32] so looking at netplan, and the syntax is very easy to follow, makes sence, BUT, how do I down/up the interface ? [18:37] Ussat: I think with ip [18:37] i.e., no new tool for that [18:37] but I could be wrong [18:44] looking now [18:48] ip link set up/down [18:49] correct [18:49] * cyphermox adds that to the FAQ [18:54] its just ifup/ifdown was...simpler [18:54] I have a test vm to play with all this, just getting prepared [18:55] well, multiple test vm's [18:55] Ussat: continued feedback is welcome [18:55] glad to help [18:56] OH, BTW, if one decides to install the ifupdown package, it will still not work, because /etc/networ/interfaces is not populated....someone may (like I did) make that assumption [18:57] BTW I am testing on server 17.10 [18:57] /etc/network/interfaces [18:57] they syntax is VERY easy to fillow and is very intuitive though [18:58] first look at it was easy to tell what was what [19:04] /etc/network/interfaces is never filled in automatically, that's always done as a "discovery" step as the installer runs (hence why it will try to do DHCP and then ask you for stuff). The installer now simply writes config to /etc/netplan instead [19:09] there's no ifupdown backend for netplan yet right? [19:09] is there a way to override netplan and force the use of the traditional /etc/network/interfaces at all? [19:09] or no? [19:12] I thought netplan just wrote the config files and then it's on you to use the existing tooling? [19:18] Honestly, I dont see a need for a ifup/down for it...it took me about...3mins to figure out what I needed [19:19] teward: that is one of the questions we will put in the FAQ as well. [19:19] nope, existing ifup./down doesnt work, use the ip command I posted earlier [19:19] cyphermox: ^ have a sneak preview answer for teward? :) [19:19] dpb1: i arrived late, but i'm asking this because it's fubaring a VPS installation :) [19:20] Honestly, it took me very little time to figure it out [19:20] and I kinda like it better [19:20] just my 2 cents [19:20] well when you are stuck using SolusVM for VPS management and it in turn hates everyone and wants the traditional method of configuring things... [19:20] sarnold: yes, that's correct. That is the intent. not to be a total abstraction, just to abstract the config generation. [19:20] you're stuck. [19:30] sarnold: precisely, it's just a translator for $backend [19:36] teward: if all you have is SolusVM, you probably can't preseed the install a special way either, so you're probably quite stuck indeed. [19:37] cyphermox: true statement, this is why I prefer mounting the ISO in KVM even though Solus manages that, so I just manually install :P [19:37] but the issue still remains. [19:38] that VPS provider should update their magic to take into account that things may not just be ifupdown or ifcfg [19:38] tell it to Solus not the provider :P [19:39] yeah, becoming a VPS is so cool we'll make so much money [19:39] vpsbasement.com [19:40] * drab now wonders if that actually exists [19:40] it's not, quick, it's a great business opportunity [19:42] drab: I claim 30% profits, it was my idea ;) [19:53] cyphermox: well i claim 60% profits because i own the domain now as a subsidiary of Dark-Net.IO, and therefore own the name unless you can prove otherwise. :P [19:54] well i could rather :P [19:54] but i'm lazy and don't want to set up *another* business. [21:17] powersj: hey, I think I'd really like to know more about how your preseed is being done for the smoketesting, because preseeding layoutcode works here, at least the way I've always learned how to do it (preseeding layoutcode, not layout) [21:18] auto=true url=http://people.canonical.com/~mtrudel/preseed/full-bionic.cfg [21:19] powersj: ^ so far, this is working with a mini.iso on ppc64el, I'm at installing the packages [21:19] cyphermox: https://paste.ubuntu.com/25956524/ is what we use for the pressed itself [21:19] my preseed isn't quite complete though, missing bits here and there to partition correctly, and to set the hostname [21:21] yeah, your preseed is basically the same w.r.t layoutcode. [21:24] cyphermox: thoughts on what to try next? have you used a full server iso? [21:25] not yet, that goes next [21:25] I'm adapting your preseed to remove anything unrelated, and I'll run one test with mini.iso and one with a full iso [21:26] but hey, I don't see why that would really matter to picking a keyboard. [21:26] ok thx [21:29] initial test done, looks fine with my old preseed [21:29] so now I'm running mini.iso with yours, but added a small change to recognize that it's different [21:29] (preseeding us:intl, which is definitely not default) === icey_ is now known as icey === fyxim_ is now known as fyxim === arosales_ is now known as arosales === kklimonda_ is now known as kklimonda [22:03] if you're on the subject of preseeding [22:03] I can't stop the installer (of mini.iso) from asking which drive to partition and which to install grub to... [22:03] im always on the subject of preseeding [22:03] any clue about that? [22:04] hurray for that, I almost sorted out all the issues except that and the iface selection [22:04] dont get me started on iface selection [22:04] yeah, I remember your qs from days ago... [22:05] cyphermox: powersj ^^^ if you have any hint [22:05] i use for grub [22:06] d-i grub-installer/only_debian and d-i grub-installer/with_other_os set to true [22:06] i never get asked for which drive [22:06] ^ same [22:06] https://raw.githubusercontent.com/powersj/ubuntu-server-preseed/master/preseed.cfg is what I've used in the past for 100% fully automated [22:07] for networking, i read an thread somewhere that ubuntu has a bug with offboard nics, even if you set the interface statically to say eth4, afterwards it will try eth0 and fall on its face [22:15] one thing I wish was more easy, adding modules to a kernel after an install [22:16] /etc/modules? [22:16] Epx998: you mean out of tree modules? [22:17] nacc: yeah an ixgbe driver again :D [22:17] Epx998: not sure why that should be 'easy' [22:18] Epx998: use dkms, i guess [22:18] more easy :D [22:18] Epx998: dkms is pretty easy [22:18] Epx998: sounds like a vendor problem, tbh. [22:19] i built a new netboot initrd.gz with the 3.13.0-117 kernel i think, with the latest ixgbe driver - once a OS is deployed, need to get the installed kernel with the same module. [22:19] dkms havent used it before let me check [22:20] Epx998: right, you want dkms, which lets you maintain a kernel module source that gets rebuilt against all installed kernels [22:20] Epx998: that's the problem dkms tries to solve [22:20] nacc: I can do this during a netboot install? [22:21] Epx998: you'd still have to build the isntaller module by hand, I think [22:21] Epx998: you could, if the dkms package was available to install [22:21] hmm [22:21] Epx998: that is, you need to install a binary package that contains the dkms module source [22:21] Epx998: `reverse-depends dkms` to see example packages [22:22] means hacking at my network installer more [22:23] top [22:24] what i have now almost works, i have ixgbe available after a restart from an install, but my post scripts are not executing [22:49] d-i preseed/run string does not run as expected [22:57] Epx998: powersj sry had to go afk, I have both options set (debian_only and with_other_os) and I still get asked [22:57] recently tried to rpeseed a desktop with 3 disks, ssd which was supposed to be root/main and 2 other spindles [22:57] and got asked [22:57] can pastebin the preseed and boot line if it helps [22:59] is your device sda? [22:59] yeah [22:59] that's what I end up putting in anyway , same for the grub question, have to type out /dev/sda [23:01] https://gist.github.com/anonymous/14cfc3ad49d230a89c4803428a8a4882 [23:01] that is one of my preseeds [23:02] better yet https://gist.github.com/anonymous/0973564df935ce9f60ade0cbc48269d1 [23:03] http://dpaste.com/2G8TKTN [23:03] that's what I'm using [23:04] I do have -> d-i partman-partitioning/no_bootable_gpt_biosgrub boolean false as well [23:04] the only "wird" thing I do is the custom partitioning, because by default the swap is craeted as 2xram which makes no sense on newer systems [23:04] there is a way to get all the options, sec i just did this last friday to get passed a kernel module question [23:05] but that's after drive selection anyway [23:05] I've seen some ppl piping in scripts to select the drive [23:05] debconf-get-selections --installer > file [23:05] that gets you every d-i option, good reading [23:05] but there seems more complicated than should be needed to write to the "first connected drive" [23:05] umhk, will take a look, I thought that was including all the installed packages etc [23:06] on --installer [23:06] good one [23:07] Epx998: I get nothing out of it [23:07] # Install the GRUB boot loader to the master boot record? [23:07] grub-installer grub-installer/with_other_os boolean true [23:07] thats from get selections output [23:08] right, and I have it in my preseed, line 68 in the pastebin [23:08] # Install the GRUB boot loader to the master boot record? [23:08] grub-installer grub-installer/only_debian boolean true [23:08] same comments added to each heh [23:09] yeah and I have that before the other one :) [23:09] line 67 [23:09] # Device for boot loader installation: [23:09] # Choices: Enter device manually, /dev/sda (ata-MB1000EAMZE_9WK3CL1J), /dev/sdb (usb-General_USB_Flash_Disk_0415100000014014-0:0) [23:09] grub-installer grub-installer/choose_bootdev select /dev/sda [23:09] do you have that in urs? [23:11] without the systemd predictable naming... was it ensurad that the disk connected to sata0 port was gonna be sda? [23:11] i do not - but... [23:11] or was that also up for grabs? [23:11] i dont get asked what youre getting asked :D [23:11] right [23:11] that's what I thought you said [23:11] so I don't get why I am... [23:11] and I don't get asked if there's only one disk [23:11] so it's clearly something happenign with multiple devices [23:11] so you have 2 disks [23:12] 3 [23:12] i only ever have 1 [23:12] but yeah, 2 will do too [23:12] ah [23:12] yeah then I don't get asked either [23:12] if i had 2, i may need that line [23:12] ok, I'll try that line [23:12] we can only break it worse [23:14] could be worse in that, your farm is running ub12 with a provisioning setup that uses both a ubuntu preseed AND a kickstart cfg file :D [23:21] like I said, it was... [23:22] even ub11 and I found one ub9 [23:22] so there, I win, I'm the king of cruft :P [23:22] did it work? [23:23] sort of, they couldn't do anything anymore, but ips were given out and traffic was making it to the internet [23:24] i meant the other fix [23:25] but i hear ya, my dev team wants an ub16 builder and said the ub12 packagesa were the same [23:25] oh, I don't know, I've added a task on my todo to try later, I just thought I'd ask since ppl were talking about preseed === ScottE_ is now known as ScottE [23:59] if i change my netboot kernel to match whats installed by the installer, would the kernel modules i build into netboot be added automatically?