[00:06] Is it possible to expand / to another disk (usb in this case) http://paste.ubuntu.com/890030/ plenty of room not on that one [00:07] not after the fact, you'd have to have used LVM to begin with [00:07] What's LVM? [00:08] it's not a good idea to do that anyway, one USB flash drive is bad enough, two is going to make it perform much worse [00:09] !lvm [00:09] Tips and tricks for RAID and LVM can be found on https://help.ubuntu.com/community/Installation/SoftwareRAID and http://www.tldp.org/HOWTO/LVM-HOWTO - For software RAID, see https://help.ubuntu.com/community/FakeRaidHowto [00:09] Its not that bad after the cache sizes gets really large :) [00:10] So there is nothing to save it now? [00:11] technically possible but very involved, quicker and easier to start over with LVM [00:12] Would it be possible install ubuntu server via a virtual machine, onto a different usb drive, then just swap them out? (the server doesn't have a monitor anymore) [01:56] hi All [01:58] i'm running ubuntu server (lucid) .. i'm trying to load an "user" script at login .. i tried with rc-update but no clue ... the script is executed as root and not in the home directory [01:59] it is a simple shell script to run a pytohn services ... i'd like to load it when the machine start (so to avoid to login with ssh and start the command from the shell manually) [01:59] have you any hints on how can i do ? [02:01] epifanio: should this python service run as root? [02:01] twb no, i need it as normal user [02:01] Then run crontab -e as that use, and write an @reboot entry into his crontab [02:02] *as that user [02:02] my server has no desktop environment ... i tried on an other linux box (xfce4 based) and actually a solution can be to copy a runscript.desktop inside the directory : /home/user/.conf/autostartup/ [02:02] oh! [02:02] i'll try it :) [02:05] to have it running at each boot .. i have to use all '' [02:06] .. all : '*' ? [02:06] like : * * * * 1 executable arg1 [02:08] oh i see .. @reboot sorry [02:13] man 5 crontab, IIRC [02:19] @reboot /usr/local/bin/ipython notebook --no-browser [02:20] works great! thanks!!! [02:25] You should probably just use a shebang [02:26] Hey guys. Not sure if this is the right place to ask but I have a firewall system with two IP on two subnets on the same LAN which is how it is supposed to be for now however I have dhcpd running which has a subnet defined for only one of those lans and I find it's mostly working however the clients are getting the dhcp config from the other lan IP even though it's the right config. For example the fw has 10.1.0.1/24 and 172.16.0.1/24 on the ... [02:26] ... same nic and it is server dhcp to 10.1.0.0/24 and it's doing it right except it's sending the data for the 10.1.0.0/24 network from the 172.16.0.1 IP. Does anyone know if there is any way to confine it to that one IP address? [02:27] jetole: slow down. You have one network running two IP ranges, e.g. 192.168/16 and 10/8? [02:27] twb: sure. yes [02:31] twb: the firewall has IP's for each net. They are both on the same broadcast LAN. The firewall is serving dhcp for 10/8 via 192.168/16 IP. 10/8 is the only configured subnet in dhcpd. The clients are getting the correct config. They are just getting it from an IP outside of the scope of the subnet they are receiving it from [02:31] twb: does that make sense? [02:42] Sorry, was elsewhere [02:42] it's cool. any thoughts? [02:43] "getting config" as in they're seeing a DHCPACK? [02:43] As in they are being properly setup with IP, routing, DNS servers etc. [02:44] As far as I can tell everything is working as it should except the server is sending it's reponses from the wrong IP but that doesn't seem to interfere. Nothing seems broken. Just logistically speaking it doesn't look right [02:44] oh I see [02:45] So what you're saying is that the server running dhcpd has (say) 10.0.0.1 and 192.168.0.1, and it's sending DHCP responses saying "you are not 10.0.0.2" but it's sending them with a src ip of 192.168.0.1? [02:46] it's sending responses saying "you _are_ 10.0.0.2" but from the source ip of 192.168.0.1. yeah. Thats about it [02:46] IIUC that's because UDP can't know which IP to use on a multi-IP iface, so it picks the widest/primary IP on that iface. If both IPs on that iface are in the same subnet, you can use "ip address" to set which is primary [02:48] well they are on different subnets. I don't think this is anything UDP specific but I think dhcpd is replying based on some logic that says just use this IP but I'm wondering if there is some way I can define which one to use [02:48] I haven't figured it out yet from the man page though [02:52] Dunno, sorry [02:52] Try talking to the dhcpd people [05:24] got a question, so i can telnet on port 22 to my server but cannot ssh.. [05:24] i am able to ssh for a few hours.. than when i come back to it.. it wont let me ssh.. it times out [05:24] any ideas? [05:41] esuave: 22 is not for ssh [05:41] Er, sorry [05:41] 22 is not for telnet. [05:41] ? [05:41] right.. but i can telnet on 22 [05:41] telnet runs on tcp/23, ssh runs on tcp/22. [05:42] i can telnet to the server on port 22 which means the port is not being blocked. [05:42] i just cant ssh in to the box. [05:42] its like the server is going into some kind of sleep mode or something [05:42] Oh, you're abusing telnet instead of using nc [05:43] It is most likely PMTUD [05:43] netcat works too [05:43] Are you blocking ICMP? [05:43] no ports are forwarded [05:43] the correct ports are forwarded [05:43] That's not what I asked. [05:43] i was able to ssh just fine like 2 hours ago [05:43] i have to reboot the machine to ssh again [05:44] Please confirm that ICMP is not blocked. [05:46] twb: ICMP would be on the router? [05:48] i would think it wouldnt be blocked since i reboot the machine and I am able to ssh? i dont know.. it seems like a few hours later it always locks up and i cant ssh :/ [05:48] its odd [05:48] esuave: on ANY router between the SSH server and SSH client, or on the SSH server, or on the SSH client. [05:49] If you can't ping the host, ICMP is probably blocked. [05:49] i can ping it [05:49] However just because ping works doesn't mean the PMTUD types of ICMP aren't blocked. [05:49] OK, have you looked at the logs of the ssh server? [05:49] Are you on a dynamic IP? Maybe the IP just changes and you don't know the new IP. [05:50] yeah.. nothin unusual there [05:50] nah its static [05:50] If you have an SSH session going and it hangs, maybe you don't have TCP keepalives turned on in .ssh/config ? [05:50] you think the computer might be doing some kind of power saving mode? [05:50] from a bios setting? [05:50] That is also possible. [05:50] One would hope the BIOS fucks off as soon as the kernel loads. [05:51] yeah cuase the server works perfect for like 2 hours than i goes blank.. [05:51] Is it always exactly two hours? [05:51] yeah [05:51] That is highly suspicious. [05:52] yeah.. i might just do a reinstall :/ [05:52] Can you reproduce this issue with other hosts on the client side, perhaps at other sites? [05:52] its just a test server anyway [05:53] nah.. thats the only host i have running on that IP [05:53] its an old piece of crap computer too [05:53] that doesnt probably help either [05:53] lol [05:54] oh well. ill try a reinstall.. maybe move to heron. Thank you for your help twb! [05:54] hardy? [05:55] If you are running something older than hardy you have big problems [05:55] All of those (except maybe 6.04 dapper) are already end-of-lifed [05:55] nah im running 10.04 right now [05:55] but i wanna move back to hardy [05:57] He's nuts. [07:20] morning === rickspencer3_ is now known as rickspencer3 === smb` is now known as smb [08:21] New bug: #959073 in samba (main) "package samba 2:3.6.3-2ubuntu1 failed to install/upgrade: subprocess installed post-installation script returned error exit status 1" [Undecided,New] https://launchpad.net/bugs/959073 [08:22] <_ruben> urgh .. stupid "kernel time sync status change" log messages, cluttering up my logs [08:25] _ruben: tell logcheck to ignore them [08:26] I gave up trying to actually get rid of them 18 months ago [08:32] <_ruben> twb: i *could* add a grep to my tail -f, but meh :) [08:34] grep --line-buffered [08:34] BTDTBTTS [08:35] tail -fn0 /var/log/syslog /var/log/auth.log | grep --line-buffered foo is one of those things I type by pure reflex [08:35] Oh add an & on the end [08:35] Nearly as much as egrep -v '^[[:space:]]*(#|$)' /etc/foorc [08:41] morning o/ [08:44] Don't call me o/ [08:49] morning all [09:02] morning jamespage, did you have a look at my last upload? [09:02] bencer_, good morning [09:02] looking at it right now [09:02] great! thanks [09:33] oh this is just great [09:35] I fixed cobbler-ubuntu-import -U to actually work, just that now it removed all systems using precise-* [09:35] and the profiles [09:43] yeah, it's totally hosed :( [09:43] it'll rename the old profile, and then later remove it [09:43] then all the systems using the old profile will get deleted as well [09:43] smoser: ^ [09:47] tjaalton: well i doubt systems will get deleted :) [09:47] koolhead11: how so? I've reproduced it [09:47] you can system edit and modify profile it should now point too [09:48] I have a distro/profile I'm not using. added a system using the profile, then deleted the iso so d-u-i would update it. after it's finished the system is no more [09:49] tjaalton: on paper systems and profiles are two separate entity i suppose. System is a wrapper on top of profile which means when you want one of your VM to say install "glance" other "nova" another "keystone" in terms of openstack [09:49] you will create 3 system file on top of the precise profile === fenris is now known as Guest97336 [09:49] yes, I don't understand how all that is relevant though [09:50] i can only think of replacing my new profile to available system. [09:50] if system gets deleted i am afraid its a flaw [09:51] no shit :) [09:52] tjaalton: :D BTW i have not touched the shiny web-interface, i do all my mighty cobbler stuff with command line [09:52] this is fairly new functionality in c-u-i that got merged recently, and apparently not tested at all, since before my changes it would just leave "tmp-$profile" and fail [09:53] now I'm thinking perhaps to revert the changes so no one will break their systems.. [09:55] jamespage: the systems bug on oneric would be fixed now? [09:57] koolhead11, cascading deleted [09:57] * jamespage sighs [09:58] dependent objects get deleted in cobbler (well they did last time I looked) [10:05] tjaalton: well your correct then. [10:06] and i would say that is a flaw in cobbler then. :D [10:13] koolhead11, tjaalton: not its best feature [10:16] found the bug [10:16] cobbler distro rename --name="$old_distro" --newname="last-$old_distro" [10:16] descendants=`cobbler profile find --distro="$old_distro"` [10:16] what's wrong with the second line? :) [10:19] tjaalton: try cobbler distro rename [10:20] "--name is required" ? [10:20] it will give you the exact comamnd parameters also its not good idea to rename distro. most thing cobbler makes sense with starts from profile IMHO :) [10:21] tjaalton: provide the current name then [10:21] I know what's wrong with the code [10:21] no need to teach me cobbler cli :) [10:21] tjaalton: sorry sir. :) [11:01] the fix is to use --distro="last-$old_distro" in the descendants= line [11:14] uploaded === EMKO|2 is now known as emko [11:27] bencer_, uploading now [11:41] jamespage: cool! so now what's next? an archive admin needs to review it from the new queue, right? [11:42] bencer_, yep - might get some feedback but hopefully they will be accepted. [11:43] bencer_, Riddell has stuck it on his list of things TODO today [11:45] jamespage: cool thanks [12:01] oh fun, now ubuntu-enlist profile is broken === bladernr_afk is now known as bladernr_ [13:22] Is there anything in Ubuntu 10.04 that somehow makes use of a file /etc/iptables.up.rules or is this something likely custom setup? [13:25] Ahhh, nevermind, it was something someone did custom, which is epic lame. [13:43] SpamapS, this bug sucks: https://bugs.launchpad.net/ubuntu/+source/psmisc/+bug/877894 [13:43] Launchpad bug 877894 in psmisc "fuser forks and never reaps its children (dup-of: 876387)" [High,Confirmed] [13:43] Launchpad bug 876387 in psmisc "fuser forking uncontrollably in cron job" [Undecided,Confirmed] [13:46] I'm having trouble accessing an apache site running on my local network. [13:50] hello guys [13:51] New bug: #959289 in nova (main) "missing nova-consoleauth init/init.d file" [Undecided,Invalid] https://launchpad.net/bugs/959289 [13:53] is it must to regestar the domain name into the domain reseller? [13:53] sorry i'm new [13:58] 4 weaks ago i install ubuntu server 11.10 into my virtual box. [13:58] after that i follow how toforge The Perfect Server - Ubuntu 11.10 With Nginx [ISPConfig 3] [14:07] jjohansen: if i ship a profile in /etc/apparmor.d/lxc/lxc-default, which gets optionally transitioned to by /etc/apparmor.d/usr.bin.lxc-start, [14:08] jjohansen: from debian/rules should i just do 'apparmor_parser -r' on that file? [14:08] jjohansen: (on the other file i'm doing dh_apparmor, but i can't do that with this file right?) [14:09] stgraber: i'ts not quite done yet, but i'm thinking of just going with lp:~serge-hallyn/ubuntu/precise/lxc/lxc-aa for per-container profiles. [14:09] hallyn: I think so, but I am no expert on dh_apparmor, lets ask jdstrand^ [14:09] stgraber: i.e., i won't create custom profiles, just support them if user wants to creat them [14:10] jjohansen: thanks [14:10] libvirt only creates them when vms get defined, and doesn't create a default, so i can't look to it for an example :) [14:11] hallyn: how are you dealing with nesting then? I thought having per-container profiles was the only way to allow /sys/fs/cgroup/*//** [14:11] hallyn: you don't want to run apparmor_parser -r from debian/rules. I assume you mean in some postinst? [14:12] stgraber: we'll just not support it by default, but offer (on wiki?) profiles which support it? [14:12] stgraber: or we can quickly add that later, i just want to focus on getting per-container support in first [14:13] jdstrand: well, IIUC dh_apparmor --profile-name=usr.bin.lxc-start from debian/rules creates the postinst bit for me [14:13] so yeah, i guess i would add the parser line to postinst myself :) [14:14] jdstrand: but i guess my q was whether there is some other dh_* i should use instead, [14:14] or if that's ok [14:15] jdstrand: and yes, if you hadn't corrected me i would've unthinkingly, wrongly put it into debian/rules :) thanks [14:16] hallyn: I am not totally clear on what you want to do. you ship /etc/apparmor.d/lxc/lxc-default, which you use dh_apparmor on, then you want to optionally support /etc/apparmor.d/usr.bin.lxc-start, which you do not ship? [14:16] hallyn: right, getting per-container profiles at all is more important than having the nesting by default, agreed [14:17] jdstrand: no, /etc/apparmor.d/usr.bin.lxc-start is entered always by lxc-start; then [14:17] jdstrand: right before running /sbin/init, it will manually aa_change_policy to a new policy; [14:17] jdstrand: by default that will be /etc/apparmor.d/lxc/lxc-default, [14:17] jdstrand: or the config file for the container can specify another policy under /etc/apparmor.d/lxc/lxc-* [14:18] jdstrand: of course i could stick clsoer to the libvirt model and always create a unique policy for each container... [14:19] hallyn: so you just need to make sure /etc/apparmor.d/lxc/lxc-* is loaded so aa_change_profile() doesn't fail. correct? [14:19] jdstrand: the reason i'm not is that there's question then about whether the general lxc-create program should do that, or the per-distro templates should [14:19] jdstrand: yes [14:20] hallyn: I don't think you can solve that wholly in packaging. ie, if the user changes the config file, the user won't fiddle with 'dpkg-reconfigure' to make sure that the policy is loaded [14:20] hallyn: seems that the lxc tools need to load the policy that is specified in the config file [14:21] (like we do in libvirt) [14:21] jdstrand: meaning on container start, do apparmor_parser ont he specified policy file? [14:21] hallyn: if the thing doing the container start reads that config, yes [14:22] jdstrand: i was just going to require users to load the policies they create... wasn'g going to try and second-guess that in packaging. only the deafult policy i want to load at pakcage install [14:22] jdstrand: then, later on we *will* create policies for *some* containers, which we'll load after creation [14:22] hallyn: that works too-- then just ignore the lxc-* stuff as well. you aren't updating it via packaging anyway- no reason to fiddle with it in packaging postinst [14:23] wait... [14:23] jdstrand: 'that works too' = loading policy at cotnainer start? [14:23] jdstrand: ignoring it [14:23] ah [14:23] either load it at container start, or just ignore it in packaging completely, letting the user do it [14:24] jdstrand: but can i load just the lxc-default profile at package postinst? [14:24] hallyn: sure. that is a file you are providing, correct? [14:24] jdstrand: mind if, when i get it working, i send you a pointer to the bzr tree with the diff? [14:24] yes [14:25] hallyn: yes, then that makes sense. I'm just saying if you aren't providing the files via packaging, don't manage them via packaging. you could make it easier on the user and load it on container start, but that is a different issue [14:25] jdstrand: ok [14:26] jdstrand: thanks, ttyl [14:26] np [14:36] jdstrand: say, i have a diff q. libvirt-bin.postrm does delgroup libvirtd. I thought I'd read somewhere that policy was to not remove users/groups at package removal? [14:36] (wondering whether to do same for lxc) [14:37] hallyn: it isn't policy. some people remove it some don't. there are arguments for and agaisnt it. it has been discussed a bit in Debian [14:38] jdstrand: ok - thanks [14:38] hallyn: the trend seems to be more people are doing it [14:38] (doing the removal) [14:52] jcastro: that bug, btw, is fixed in precise [15:01] New bug: #959352 in lxc (universe) "Ephemeral containers have "/rootfs" prefix in /proc/self/maps entries" [Undecided,New] https://launchpad.net/bugs/959352 [15:04] SpamapS, doing wonders for me over in what we ship for real. :) [15:05] jcastro: marked that bug for SRU.. [15:05] jcastro: psmisc's fuser is apparently pretty broken in 11.10 [15:31] New bug: #959393 in samba (main) "samba deb archives should use lzma or bzip2 compression" [Undecided,New] https://launchpad.net/bugs/959393 [15:34] jdstrand: im just trying to get updated did anything happen with the keystone MIR last week [15:36] zul: yes, I conditionally ackd it [15:36] zul: see the bug (881464) [15:37] jdstrand: cool thanks [15:37] zul: welcome back! [15:37] jdstrand: thanks [15:43] hallyn: I want to alert you to the existance of bug 959352; it shouldn't cause too many people problems, but what problems it does cause will likely be hard to diagnose [15:43] Launchpad bug 959352 in lxc "Ephemeral containers have "/rootfs" prefix in /proc/self/maps entries" [Undecided,New] https://launchpad.net/bugs/959352 [15:44] benji: yeah, i'm looking at that right now [15:44] cool [15:47] jdstrand: have you started the horizon security review yet? [15:50] zul: nope. I am working on juju, then cobbler then horizon. that was the priority as specified by Daviey et al [15:51] jdstrand: gotcha thanks :) [15:52] jamespage: btw, i've just pushed to lp and proposed a merge of a changeset that includes a zentyal chapter on the ubuntu-server guide :) [15:52] bencer_, nice one! [15:52] i would have liked to include some screenshots [15:53] but i didn't see any screenshots on the server-guide and despite i asked on the ml and on irc channel, nobody replied [15:53] so i dont really know the procedure or even if its allowed [16:01] New bug: #959426 in nova (main) "nova services start before mysql on boot" [Undecided,New] https://launchpad.net/bugs/959426 [16:19] jamespage: you have a few minutes for some questions about the jenkins-slave packages in precise? [16:19] ahs3, sure [16:20] jamespage: groovy. so, the upstart attempts to download slave.jar and ends up with a zero length file -- i'm assuming i'm doing something stupid, yeah :)? [16:21] ahs3: is the jenkins server running when it tries to start? [16:22] jamespage: well, i've got the master up and visible in a browser; is there another process to be run? [16:25] jamespage: the other curiosity is that /var/run/jenkins doesn't seem to get created for me automagically... [16:36] New bug: #959461 in quota (main) "Language incongruence making edquota -t" [Undecided,New] https://launchpad.net/bugs/959461 === koolhead17|away is now known as koolhead17 [16:43] ahs3, no that should be sufficient [16:44] * jamespage goes to take a look [16:44] jamespage: hrm. getting a 404 on the wget in the upstart script... [16:46] ahs3: have you configured a slave in the jenkins master for this jenkins slave? [16:46] I think it will 404 if you have not [16:55] jamespage: yup, the node is set up as a slave, and the master tries to contact it [16:55] ahs3, that has a few contradictory configurations in it [16:56] ahs3: can you walk me through what you have done so far? [16:57] jamespage: argh. found it. JENKINS_URL was not set properly in defaults :( [16:57] ahs3, that would be it [16:57] you can run slaves in a few different ways [16:57] * ahs3 reminds himself not to edit scripts late at night... [16:58] right. this was just my first attempt at setting up a slave [16:59] part of what i wanted to experiment with was the different ways to run them [17:04] ahs3, so the jenkins slave package is really aimed at [17:04] 1) deploying slaves via SSH - it has all the right dependencies to support SSH access, Java, users etc... [17:04] who here loves nscd? :) [17:04] that requires no setup on the slave - just configuration in Jenkins [17:05] ahs3: 2) where you can't SSH to the slave - you can configure the JENKINS_URL for the jenkins-slave upstart config and have the slave initiate the connect to the master [17:07] jamespage: ah, okay. so in the default case, i do no config on the slave at all, just install the packages? [17:07] jamespage: sorry, default meaning the "deploy via SSH" case... [17:07] ahs3, yep - you will need to installs the ssh-slaves plugin on the Jenkins master [17:07] (I'm working on making that part of the default jenkins install) [17:08] jamespage: nod. k, i was doing more than needed [17:25] Hi, which cloud-image works best with LXC? precise-server-cloudimg-amd64.tar.gz -or- precise-server-cloudimg-amd64-root.tar.gz -or- precise-server-cloudimg-amd64-disk1.img (all from cloud-images.ubuntu.com) [17:25] med: precise-server-cloudimg-amd64-root.tar.gz [17:26] utlemming, danke [17:26] med: you can automagically use it with "lxc -n AutomagicCloudImage -t ubuntu-cloud" [17:26] er...lxc-create [17:26] and there are options to uncloud them too [17:33] utlemming, I'm doing this inside of openstack so I can't use the automagic. [17:53] should i be using lvm if i am virtualizing with hyperv? === blendedbychris1 is now known as BlendedByChris [17:56] BlendedByChris: up to you [17:56] just curios if there is an advtage [17:56] BlendedByChris: use lvm because you want/need to, nothing to do with the virtualized host [17:56] no more / less than using lvm on a physical host [17:59] well my linode says i can't resize the disk with lvm [18:03] just wondering if that's a xen thing or hyper-v thing or what [18:09] Daviey: ping [18:10] zul: ping is so outdated... i use spark-ping.. written in ruby and everything [18:10] Daviey: heh so ok webob... [18:11] $ spark-ping ubuntu.com [18:11] ^C▅▆▂▄▅▆▇▃▄▅▆▃▄▅ 96.6 ms [18:11] Daviey: Error: "C▅▆▂▄▅▆▇▃▄▅▆▃▄▅" is not a valid command. [18:11] Response.request and Response.environ got deprectated in 1.1.1 which is in precise [18:11] I've tried to read the man page I swear, I am grabbing stuff off of rsync remotely, to a server at my house, and when the ssh to my local server dies, the connection stops, how can I prevent that? [18:12] and all of openstack uses it *everywhere* [18:12] glance, keystone, nova [18:12] hilarie: run it in screen [18:12] there is no screen hooked up to the server at my house :( [18:12] Daviey: and it got undepracted in 1.2b1 but with a slight variation [18:12] zul: is support removed, or just noisy with depreccation warnings? [18:13] its just noisy deprecation warnings [18:13] hilarie: no, run "screen" command.. it gives you a persistence console [18:13] hilarie: then reconnect later on with 'screen -r' [18:13] zul: is it would just patching out the warnings? [18:13] s/would/worth/ [18:14] Daviey: yeah might be [18:14] Daiviey what you are saying is read the man page on screen :) got it [18:14] zul: seems less destructive than a new upstream version at this stage. [18:14] Daviey: my thoughts exactly [18:14] zul: Might be worth raising an upstream bug, so we can track this for next release. [18:14] * ogra_ grins [18:14] stop confusing the poor bot [18:14] (it doesnt speak ruby apparently) [18:14] Daviey: ack [18:14] Although, 'undepreciated' concerns me. [18:15] ogra_: heh [18:15] Daviey: lemme pull up the commit [18:15] New bug: #959419 in postfix (main) "package postfix 2.7.0-1ubuntu0.2 failed to install/upgrade: sous-processus nouveau script pre-installation tué par le signal (Relais brisé (pipe))" [Undecided,New] https://launchpad.net/bugs/959419 [18:17] Daviey: https://github.com/Pylons/webob/commit/627593bbcd4ab52adc7ee569001cdda91c670d5d [18:20] zul: in that case, it makes total sense to remove the deprecation warnings. [18:21] with that patch above? [18:22] zul: do you know what upstream are QA'ing against? [18:22] depends on the project [18:22] zul: might be better just to create a patch which removes the dep warnings.. rather than morphing the characteristics [18:22] Daviey: gotcha [18:23] Daviey: but i think they are standardizing on 1.0.8 [18:38] zul: do you have a good example of a package that handles upgrades (db migrations and stuff) [18:39] not really off the top of my head [18:40] anyone else having issues starting lxc containers in precise? (lxc-start: Permission denied - failed to mount 'proc' on '/usr/lib/lxc/root//proc') [18:40] roaksoax: eucalyptus [18:42] roaksoax: http://pb.daviey.com/Viat/ [18:42] obv. it should be using invoke-rc.d [18:43] Daviey: yeah I saw.. though in our case is not that helpful because we would likely have to update it with every single new bzr branch that we upload [18:45] roaksoax: right.. well, it's safe to do a south migration on each update. [18:45] is us.archive.ubuntu.com usually pretty slow? [18:45] roaksoax: if there is nothing to do, then it'll handle it ok [18:46] Daviey: right, postinst only handle configure|reconfigure [18:46] Daviey: but i'll figure it out [18:47] roaksoax: every package gets configured, regardless if it is an upgrade or not :) [18:47] Daviey: so on first install, if the dbconfig-common question has been answered true, then it does syncdb and stuff, if its answered as NO, then it does not run syncdb, but on every uipgrade you need to run it [18:47] roaksoax: $1 = "upgrade" btw [18:48] Daviey: http://bochs.sourceforge.net/cgi-bin/lxr/source/build/debian/postinst.ex [18:48] Daviey: postinst should not handle 'upgrade' [18:48] ah, that is preinst.. crap [18:49] yep [18:50] roaksoax: The other thing you can do, is run migrate on the upstart job? [18:51] roaksoax: nova essentially runs migrate on each start. [18:51] Daviey: uhmm ok, I'll take a look at it [18:52] roaksoax: But really, why not run migrations on each configure? [18:53] Daviey: migrations are run on each configure, the different is if you install for the first time, and the DB is not created by dbconfig-common, then it will fail and cause a install failure [18:55] ah [18:56] New bug: #959597 in puppet "missing parenthesis in process_name.rb produces ruby warnings" [Undecided,Confirmed] https://launchpad.net/bugs/959597 [19:14] any reason why hyper-v would be masively slow [19:15] gary_poster: whatever happened with lxc-start-ephemeral vs. overlayfs? [19:15] gary_poster: i was expecting to find aufs had been re-enabled, but it doesn't seem to be? [19:17] hallyn, hi. I want to use overlayfs, and am trying to use it as long as I can. My comment to the list was hot on the heels of finding another issue--that we have a workaround for and that Andy had a patch for last I looked. aufs is my backup-plan--and a mediocre one at that, since we may indeed discover issues with it as well, because we have used overlayfs since January. I'd like to see [19:17] an option in ephemeral to use aufs [19:17] it has not been a priority, and overlayfs hasn't bitten us lately [19:18] but I should get someone to do it, unless you want to, hallyn. [19:19] gary_poster: well, if it's been working lately... i just couldn't remember where we'd left that [19:20] what was the thing that was failing again? [19:20] hallyn, the hard link of a 0444 file broke [19:21] chmod 444 I mean [19:21] hallyn, when's the freeze? Has it already snuck up on me? I know it is soon. [19:22] not yet anyway [19:22] heh [19:22] checking [19:23] another freezer mar 22, though i don't think that's the last one [19:23] jinkeys, i'd better get these other things in then [19:23] btw hallyn did you see bug 959352? [19:23] Launchpad bug 959352 in lxc "Ephemeral containers have "/rootfs" prefix in /proc/self/maps entries" [Low,New] https://launchpad.net/bugs/959352 [19:23] eh you probably triaged it [19:23] yeah [19:23] sorry :-) [19:24] oh yeah haven't had a chance to reproduce that yet (the machine i was using was failing lxc-=start-ephemeral entirely so i couldn't reproduce to mark it confirmed yet) [19:24] ok lemme get this lxc-shutdown thing wrapped up and then i'll look at ephemeral [19:24] cool tahnks hallyn [19:25] gary_poster: I guess think it through, and if you think you want aufs re-enabled, pls file a bug today or tomorrow and i'll re-enable it [19:25] ack hallyn will do [19:25] thanks - ttyl [19:52] hallyn, I finally got an answer about bzr + git + quilt, and it is not a particularly good one. In sum: (1) there is no good bzr + git + quilt answer now, according to people who should know; (2) there is a nascent bzr + quilt answer that is in the works, which has been usable recently (though is not necessarily right now), and you could check with Jelmer Vernooij if you wanted to track that or learn more. [19:53] gary_poster: perhaps this should be a tutorial/ahckfest at uds [19:53] (He's Canonical, on the bzr team, nick "jelmer") [19:54] hallyn, sounds very reasonable. This sounds like the problem is affecting a decent number of people [19:54] gary_poster: it just means i merge/sync by hand rather than being able to have bzr do it. it'd be neat to have a better sol'n, but it's not really in the way [19:54] fair enough [19:56]