=== markthomas|away is now known as markthomas === kickinz1 is now known as kickinz1|afk === kickinz1|afk is now known as kickinz1 === zz_DenBeiren is now known as DenBeiren === kickinz1 is now known as kickinz1|afk === collizio1 is now known as collizion === zz_DenBeiren is now known as DenBeiren === lea_ is now known as lea === kickinz1|afk is now known as kickinz1 === lea_ is now known as figment [08:54] rbasak, morning - sage wants to discuss apache in 12.04 (and 14.04) with the proposed mod-proxy-fastcgi switch upstream in ceph tomorrow afternoon [08:54] are you around? [10:08] Good morning. === Lcawte|Away is now known as Lcawte === Lcawte is now known as Lcawte|Away === martinst is now known as martins-afk [13:00] sahara package is not available for ubuntu [13:00] is it? [13:01] http://docs.openstack.org/juno/install-guide/install/apt/content/sahara-install.html << As per the doc === martins-afk is now known as martinst === Lcawte|Away is now known as Lcawte [13:57] mkswap: /var/swap.img: warning: don't erase bootbits sectors on whole disk. Use -f to force. <--- should I be concerned? [13:59] rbasak, https://launchpad.net/ubuntu/+source/mysql-5.6/5.6.23-0ubuntu1 [13:59] that's a big changelog entry :-) [14:01] jamespage: did you have any changes for nova b2? [14:01] zul, nope - I rebased my power sync patch upstream [14:16] jamespage: :) [14:17] jamespage: it fails dep8 though. I still need to look at it. [14:17] (it passed locally :-/ [14:21] can you help me please? [14:22] zul, lets see if we can unstick proposed today [14:22] jamespage: sure [14:22] zul, I think nova/eventlet is the primary cause [14:23] jamespage: just uploaded a new nova [14:23] zul, I need something mindless todo so I'll work through some of the oslo. -> oslo- changes as well [14:23] at least in deps - not in core openstack packafges === Lcawte is now known as Lcawte|Away [14:51] smb: hey - have you looked in detail at the reply to "[libvirt] [PATCH/RFC] Add missing delta from Ubuntu to apparmor profiles" ? [14:53] hallyn, No, not yet. It was so long ago that I need to read up a lot. Which I thought of maybe doing next week when I finished up things I want to get done first === pgraner-afk is now known as pgraner [14:54] arges, just a reminder that horizon is uploaded for juno [14:55] coreycb: Thanks. I didn't check when I EODed yesterday [14:56] smb: ok thx [15:01] coreycb: ok done [15:02] arges, thanks! [15:28] hallyn, around ? [15:28] question on bug 1418221 [15:29] smoser: here [15:29] i dont' care, you want me to add qemu-kvm link on ppc? [15:30] it seems that a link of 'qemu-kvm' -> qemu-system-ppc64 would reportedly fix this. [15:30] what do you think ? [15:30] i wouldn't do it for ppc64 only. it would make sense for that link to point on all arch to "the default qemu" [15:30] and apparently one such thing exists in fedora [15:30] ? [15:31] well qemu-kvm package exists on ppc64 and ppc64el, [15:32] what does it do ? [15:32] ifneq ($(filter $(DEB_HOST_ARCH),ppc64 ppc64el powerpc),) install -D -m755 debian/kvm.powerpc debian/qemu-kvm/usr/bin/kvm [15:32] endif [15:33] what happens when you install qemu-kvm on ppc64el? [15:33] and who's the asshat who thought it should be ppc64le in some places and ppc64el in others [15:34] dpkg arch is el. kernel arch is le. [15:34] you have to blame debian for selection of that name for dpkg arch. [15:34] is it some kind of meta endian joke to have those different arch names? [15:35] it is. [15:35] fun times [15:35] "hey lets be funny by making pain for people!" [15:35] jamespage: hey there shouldnt be a nova-compute.upstart.in anymore should there? [15:35] haha [15:38] beisner, so how does qemu-system-ppc end up on a nova-compute node ? [15:38] is that the nova-compute charm that installs it? it looks like maybe we might be able to just have it install qemu-kvm also. [15:38] jamespage, ^ is that appropriate? [15:39] smoser, i don't have the ppc64el compute node up atm, but will in a while. then i can check into when/where it enters the picture. [15:42] ok. thanks. [15:44] zul, there should be yes [15:44] conditional start on libvirt [15:44] jamespage: ok just checking [15:45] smoser, I think its a dep from nova-compute-kvm [15:45] zul, ^^ - might be another change for you to include [15:45] but don't hold for it [15:46] smoser: ok if it turns out kvm is not being installed by qemu-kvm package pls let me know === kickinz1 is now known as kickinz1|afk [15:53] looks like the package nova-compute-kvm suggests qemu-kvm [15:53] so it doesnt get installed by default [15:54] hm.. [15:56] if we're going to continue to let libvirt depend on a symlink/shell-wrapper rather than more explicitknowledge of ppc64el, then we will need some way to ensure that users dont install libvirt and qemu-system-ppc64 without getting that link. [15:58] smoser, fyi just kicked off ppc64 deploy; will touch base soon. [15:59] it really seems like just installing qemu-kvm will do it. [15:59] so try that first (instead of symlink) [16:16] ack [16:27] coreycb, zul: taskflow uploaded [16:27] 0.7.0 [16:27] coreycb, zul: something in pypi has changed which means uscan and watch files are broken [16:27] its really annoying [16:30] zul, I see this as well [16:30] W: cinder source: virtual-package-depends-without-real-package-depends build-depends-indep: python-oslo-config [16:30] I thought the last fix I did for oslo-config should resolve that but apparently not [16:30] ergh [16:34] hey zul or jamespage whats the ETA on Juno.2 in UCA? [16:34] mfisch, we are working through it now [16:34] mfisch, probably all in by the end of next week [16:34] thanks [16:34] mfisch, quite a bit todo still [16:35] mir's and stuff will block mostly [16:35] we'll plan on moving to it when it lands [16:35] thanks jamespage [16:35] Openstack was forced to use mir also? ;) [16:35] sort of but also sort of have a Get Out of Jail Free card [16:36] I guess if it brought in a dep that really needs to go NOT IN the UCA, that peice would have to be MIRed === jgee_ is now known as jgee [16:37] not the space station, not the display, the MAIN INCLUSION REQUEST [16:37] * mfisch was bringing up memories of being on the desktop side of things at Canonical [16:39] med_, mfisch - you guys wanted openstackclient in the UCA right? [16:39] its there now but under the same terms as universe [16:39] yep thats great [16:39] we're fine unsupported on it [16:39] we're about to annotate that archive in the same way as ubuntu as to what's supported by canonical and what's not [16:40] mfisch, good-oh [16:40] perfect [16:40] jamespage, many thanks! [16:40] what about labels for the repos? [16:40] Label: filterable in pinning [16:40] (in the Releases file) [17:15] hey jamespage trad511 has some OVS questions [17:16] hallyn, having libvirt call kvm (or qemu-kvm) rather than appropriately named doesnt work. [17:16] http://paste.ubuntu.com/10190884/ [17:16] due to apparmor denial [17:19] jamespage: having a need in openstack ovs to set interface mtu, do l3 configs etc and wondering if the ubuntu /etc/network/interfaces ovs support is the "proper" way to do it. also using puppet vswitch module through puppet-neutron [17:23] frediz, leitao: http://dep.debian.net/deps/dep8/ [17:23] frediz, leitao: http://dep.debian.net/deps/dep3/ [17:25] frediz: 1.4.0~rc1 < 1.4.0 [17:26] s/-rc/~rc/ === liam_ is now known as Guest13534 [17:30] smoser: oh, interesting. [17:30] arges: ^ [17:31] hallyn: whats interesting? [17:31] arges: the powerpc kvm script can't ever have worked bc of apparmor [17:31] well, not through libvirt [17:32] hallyn: yea i've usually modprobed manually [17:32] no, i mean /usr/bin/kvm (kvm.powerpc) [17:32] it calls awk which libvirt doesnt' allow qemu to do [17:32] hallyn: wierd. i swear libvirt on power8 was working for me last i tried it [17:32] trying to think whether we shoul dhave an apparmor sub-profile for kvm or not [17:33] arges: using libvirt? [17:33] hallyn: yea [17:34] so, we could update kvm.powerpc to use only shell-isms [17:35] i dont even know if that is sufficient, hally. [17:35] hallyn, i [17:36] or just add /bin/awk rmix, to the libvirt-qemu template [17:36] i tried a while ago, and i think even the exec means loading a executable which wasnt in the profile that is crafted based on the xml. [17:36] smoser: that can't be, we do that now in /usr/binkvm [17:36] oh?hm.. [17:37] yeah the libvirt-qemu template grants exec permission to /usr/bin/qemu-system-ppc{,64} [17:38] hallyn: fwiw this is how I've been doing power8 kvm testing http://pastebin.ubuntu.com/10191161/ [17:38] sudo ln -sf qemu-system-ppc64 /usr/bin/qemu-system-ppc64le [17:39] ^ that seems to me to maybe be the best thing at the moment. [17:39] smoser, ok, we have a freshly deployed enviro up, ppc64el compute node w/ trusty hwe-u + kilo-proposed. [17:39] qemu|libvirt|kvm dpkg info to start: http://paste.ubuntu.com/10191119/ [17:39] arges: yes but uvt-kvm starts it as /usr/bin/qemu-system-$arch, not as /usr/bin/kvm [17:39] hallyn: then that makes sense why i wasn't hitting the issue [17:40] beisner, oh fiddle. [17:40] i forgot you had to use my ppa to get a 1.2.12 [17:40] as cloud archive is behind [17:40] smoser: so if you add /bin/uname rmix and /bin/cat rmix to /etc/apparmor.d/abstractions/libvirt-qemu, does that work? [17:42] maybe. but i wouldnt htink you'd want to add cat. [17:42] why? [17:42] oh i meant awk [17:43] i wouldnt want to addeither [17:43] feel free to update the scrip tinstead :) [17:43] mr sh [17:43] i guess its nto that bad. [17:43] i wouldn thtink youd want to add them as they're an attack vector [17:43] awk we could do with IFS< uname i don't know [17:43] awk for sure can arbitrarily execute code [17:43] would seem non-ideal from apparmor perspedtive, no ? [17:44] smoser, shall i install these, should be pre-patch: https://launchpad.net/ubuntu/+source/libvirt/1.2.12-0ubuntu2/+build/6955199 [17:44] it'll remain confined [17:44] right. [17:44] any recommendations for a cheap hoster, where i can pay by bank transfer (from germany)? [17:44] is there some offtopic-channel where i should ask this question? [17:45] thor77, you mgiht get an answer, you might not. i dont know. [17:45] Smark: k [17:45] smoser: k [17:45] vultr maybe , digital ocean, but i'm sure you cnsidered those. [17:45] (i have no idea on bank payment from germany) [17:45] beisner, well , those are built for vivid. [17:45] sure enough! [17:46] so they're not going to be nice. the value of my ppa was in gettin getting the cloud-archive backport for trusty. [17:47] and kilo-staging in ca has 1.2.12-0ubuntu3 - which has the patch we want to not-try, correct? [17:48] smoser: both doesnt provide pay by bank transfer :( [17:48] beisner, well, that and it doesn't (staging) build ppc64el [17:48] so thats not helpful :) [17:48] oh fiddle indeed then. [17:48] beisner, i'll build some. [17:49] smoser, ok thanks - just holler. will leave that deployment untouched. [17:56] almost too many cpus for my screen. http://i.imgur.com/LHeOa3u.png [18:25] hallyn, http://paste.ubuntu.com/10191769/ [18:25] that is pure shell, but ditches the uname -m usage. which id ont thikn you want. [18:26] beisner: My favourite part of that is the swap; that'll make all the difference. ;) [18:39] beisner, ok. so there are some debs in ~smoser/libvirt-builds on lescina [18:39] smoser: what does /proc/cpuinfo look like o npower? [18:40] eh, i won't second-guess you :) [18:40] http://paste.ubuntu.com/10191988/ [18:41] oh. and hm.. [18:41] that is possibly something you dont want to parse with shell rpeatedly [18:41] oh. but not so bad. [18:41] as we get out early. [18:42] but the uname stuff is lost. [18:42] you were making decision based on kernel uname, an i dont know a good way to do that without using uname [18:43] Gawd... So bored! [18:43] smoser: ok, i'm going to switch to yours then, thx [18:43] what about uname thoug h? [18:44] oops. and need trilaing ';' on that last case line [18:44] I see a trailing ; [18:45] smoser: well, i'm not sure 'uname -m' was the right thing ot use anyway [18:46] i.e. run uname -m under 'i386' [18:46] it really should have 2 ';'; [18:46] er.. 2: ;; [18:46] oh [18:46] for case [18:49] arges: http://paste.ubuntu.com/10191769/ does that look ok to you? [18:50] arges: btw, i've been assuming you came up with th eoriginal kvm.powerpc. maybe i'm wrong? did infinity do it? [18:50] hallyn: i didn't do it [18:50] sorry then :) [18:50] mwhudson: \o [18:51] hallyn: so why would -ppc be the default instead of -ppc64? [18:51] seems like ppc64 hardware is the most common [18:51] dnno, for safety? [18:52] i guess ppc can be run on 64-bit kernel.. [18:52] * hallyn hops over to ubuntu-devel [18:53] hallyn: i guess i would have been lazy and just relied on grep instead of doing the read line loop [18:54] arges: grep not allowed by apparmor [18:54] but anyway smoser doesn't whimp out on leg day :) [18:54] hallyn: hmm... i might need to modify my virt-detection script then [18:55] isn't that only for the init script? [18:55] that doesn't run under apparmor [18:55] ah ok [18:56] i dont know about qemu-systme-ppc being default. i [18:57] i'd be fine to have ppc64 the default. unless theres some obvious system thing. that says otherwise. [18:57] i really have no idea what that would be [18:57] hallyn, what would you think about a symlink qemu-system-ppc64le to qemu-system-ppc64 [18:57] smoser i've got some other things to push so for now gihtub.com/hallyn/qemu #ubuntu-dev has your current script. we can update it there (i'll rebase later) as we ahve better ideas [18:57] smoser: i have no idea [18:58] that would also fix our problem [18:58] seems fine [18:58] will there ever be a separate binary? [18:58] it doens't seem like the direction [18:58] ie, i dont knwo why there'd be a different binary [18:58] when they're emulating the same system. [18:58] ie, there really isnt a "ppc64el" hardware, and qemu is a hardware emulator [18:58] ok, i'll add the link [19:03] hallyn, just to be clear, the link would be qemu-system-ppc64le -> qemu-system-ppc64 [19:03] "le" not "el" [19:03] so we dont fall victim to that funny joke. [19:05] smoser: I added "usr/bin/qemu-system-ppc usr/bin/qemu-system-ppc64le" to qemu-system-ppc.links [19:06] wait. thats wrong. [19:06] oh? [19:06] $ ls -l /usr/bin/qemu-system-ppc64le [19:06] lrwxrwxrwx 1 root root 17 Feb 12 19:00 /usr/bin/qemu-system-ppc64le -> qemu-system-ppc64 [19:06] bc that should be 64? [19:06] thats what we want. [19:08] s oi'm not sure if i ca nhave the .links entry just be "qemu-system-ppc64 usr/bin/qemu-system-ppc64le" or if the first should be prepended by usr/bin anyway [19:09] but i'll do a test build [19:09] after i add arges' other fix for 1414153 === Lcawte|Away is now known as Lcawte [19:17] Odd_Bloke, you'd be surprised. I actually made it go swappy with 250-some instances fired up. === liam_ is now known as Guest73606 [20:18] beisner: :D === kickinz1|afk is now known as kickinz1 === markthomas is now known as markthomas|away [21:22] hallyn, did you drop my hack patch in your upload ? [21:30] smoser: which upload? [21:31] oh, no, i'm still test-building. should i drop it? [21:31] em [21:31] oh that's libvirt [21:31] no it's still in there. so do yo uwnat me to do another upload droppin ghtat? [21:35] hallyn, yeah, get rid of that patch. i never really intended for it to be included. [21:38] ok [21:38] sorry :) [21:39] the symlink should fix it? [21:41] smoser: look ok? lrwxrwxrwx 1 ubuntu ubuntu 17 Feb 12 21:32 qemu-system-ppc64le -> qemu-system-ppc64 === markthomas|away is now known as markthomas [22:13] looks good, hallyn === FreezingAlt is now known as FreezingCold [22:52] hello, do you know any solution for setting up a dns name based ftp reverse proxy? [22:55] nch_: dns protocol doesn't include any server names, that's why there's no name-based ftp virtual hosting like you see with http [22:59] sarnold: so it is impossible? :( [22:59] nch_: I've never seen it any done before.. [23:00] nch_: ftp is a terrible protocol anyway; what are you trying to accomplish? there's probably something better out there anyhow [23:02] sarnold: I´m using pfsense for securing my internal environment and I have two webservers (one linux and one windows), and on both my customers can upload their content with ftp, so I want a reverse proxy for different internal ftp targets with one public ip, like a reverse http proxy [23:04] nch_: sftp over ssh ought ot let you get much the same thing done without using multiple IPs === Lcawte is now known as Lcawte|Away