[00:32] anyone else have broken colorspace issues on video files using nvidia on karmic? [00:33] how broken? i know there is an issue with totem and nvidia users having to adjust the hue when they first run it [00:40] directhex: I used to have weird colour issues when viewing flv/youtube videos in mplayer/totem/xine earlier in the cycle, although after a fresh install, that went away. [00:40] chrisccoulson, adjust it where? the weird thing is it's fine for the duration of having nvidia-settings running [00:41] directhex - that's wierd. i had to adjust the hue slider to the center in totem preferences when i did a fresh install [00:41] hm, not anymore. worked last week [00:42] wow, wait, what? chrisccoulson, well spotted! why is that randomly so low? [00:42] directhex - that's likely the same issue as bug 395476 [00:42] Launchpad bug 395476 in nvidia-graphics-drivers-180 "nvidia sets HUE to -1000" [Undecided,Confirmed] https://launchpad.net/bugs/395476 [00:43] low priority? [00:44] heh, i'm not sure who set that [00:44] YokoZar i guess [00:55] dailies up in ~ 1 hr? [01:00] Keybuk: bug 452503 -- both "file" and older "blkid" worked correctly. [01:00] Launchpad bug 452503 in util-linux "blkid fails to recognize swap partition / file" [Undecided,Confirmed] https://launchpad.net/bugs/452503 === freeflyi1g is now known as freeflying [01:01] kees: did older vol_id work correctly? [01:01] actually, the argument is that file is working *in*correctly [01:01] or blkid should say "LUKS" but it doesn't [01:01] blkid should say neither [01:01] it has BOTH metadata [01:01] the main point is that it _is_ a valid swap partition. the kernel can use it, etc [01:02] it could be either, you can't pick one without risking corrupting the other [01:02] yes, but blkid doesn't know it's *not* a valid LUKS partition [01:02] that's the point [01:02] ok, that's reasonable. [01:02] it'd be nice to have a --verbose :) [01:02] if blkid tells you it's swap, and you really meant LUKS, you just trashed your LUKS [01:02] if blkid tells you it's LUKS, and you really meant swap, you just trashed your swap [01:02] now [01:02] replace LUKS and swap in those with [01:02] say [01:02] ext4 and vfat [01:02] * kees nods [01:02] (which are the other common pairs) [01:02] that's why this is a Won't Fix [01:02] almost seems like a bug in mkswap then... [01:03] indeed, I'm surprised mkswap didn't wipe [01:03] that is certainly a bug [01:03] especially given ff3bed806863d1c2075d0efda70b39ea6af9ecba [01:03] which is the code in mkswap to explicitly wipe things out [01:04] * kees changes the bug title's name [01:04] er, bug's title [01:05] quick question, what VCS are you guys using? [01:06] AdamSchackart: all kinds but mostly bzr is what you'll find in Ubuntu [01:06] AdamSchackart: bzr and git [01:06] cool [01:08] Keybuk: hey - upstart job question [01:08] Keybuk: http://people.canonical.com/~mathiaz/eucalyptus.conf [01:08] Keybuk: ^^ will the opts variable be passed to the exec? [01:08] mathiaz: no. [01:08] Keybuk: ah! [01:08] Keybuk: how should this be done? [01:08] script [01:08] mdz has some weird things elsewhere to output that stuff to a file, then source the file in the exec line [01:08] opts=foo [01:09] exec $opts [01:09] end script [01:09] alternatively just put the whole lot of pre-start and exec into one script [01:09] as ion just demo'd :p [01:09] Keybuk: great - thanks [01:09] ion: thanks as well [01:10] holy crap, I just read the moblin sreadahead pack [01:10] and really wish I hadn't [01:10] this thing really is SSD optimised ;) [01:10] find / | make-pack [01:11] Hehe [01:11] oh, I see [01:11] then that mincore()s it [01:11] so it's not quite as crazy [01:11] Ah [01:12] Keybuk: so if you call 'stop' in a script instead of a pre-start script, does it still work? === TheMuso` is now known as TheMuso [01:14] (i.e., the only thing you don't want to do in a script is fork another process that's not the one to be supervised?) [01:17] Keybuk: exec . /var/run/eucalyptus/eucalyptus.options && eucalyptus-cloud $opts [01:17] Keybuk: would something like this ^^ work? [01:17] Keybuk: (provided that the opts are dumped correctly in /var/run/eucalyptus/eucalyptus.options in the pre-start part of the job) [01:18] is kees likely to be about this evening? [01:19] chrisccoulson: he is! :) [01:20] mathiaz: a) should be in a script instead; b) really would be better to not need the tmp file [01:20] slangasek: ok [01:21] kees - i'm confused - i don't see you on the list of users ;) [01:21] mathiaz: I think the current script has a bug anyway, because calling "exit 0" from pre-start doesn't stop the job from running, which I think is what the goal was there? [01:21] slangasek: should I exec in the script? [01:21] must be empathy playing up! [01:21] mathiaz: yes [01:22] kees - i have a patch for the screensaver issue. not sure if you want to take a quick look at it (or test it)? [01:22] it seems to be working ok here [01:22] chrisccoulson: sure! [01:23] kees - the patch is here: http://pastebin.com/m239703a2 [01:23] mathiaz: do you want the job to fail, or just not stop, if eucalyptus is not configured yet? [01:23] slangasek: hm - what does "just not stop" mean? [01:24] mathiaz: it means "just not start" :-) [01:24] mathiaz: modulo the answer to the above question, I think what you really want is this: http://people.canonical.com/~vorlon/eucalyptus.conf === TheMuso` is now known as TheMuso [01:25] slangasek: exit 1 will make the job fail? [01:25] mathiaz: if you want the job to stop instead of failing, you can replace 'exit 1' with 'stop'; and then you can also mark the job as 'respawn'... [01:25] mathiaz: yes [01:25] slangasek: exit 0 will make the job not start ? [01:25] no [01:25] chrisccoulson: nice! [01:26] mathiaz: exit 0 called from the pre-start script just says "the script is done, exited successfully" and the job will still start [01:26] chrisccoulson: can you attach that to the bug, and I can get it built and get a freeze exception. [01:26] slangasek: what does script+respawn do then? [01:26] slangasek: it reruns the script until it's successfull? [01:27] mathiaz: 'exec foo' is shorthand for 'script exec foo end script' === norly is now known as ejat [01:27] mathiaz: for a normal job, the definition of "successful" is "running" [01:28] mathiaz: "respawn" says "if the job dies for any reason, respawn it" [01:28] kees - yeah, no worries. i can push it to ubuntu-desktop bzr if you're happy with the change [01:28] slangasek: ok [01:28] mathiaz: updated the script a little, to use the Keybuk-approved 'stop' runes [01:28] chrisccoulson: if it works, yeah! :) [01:29] slangasek: { stop; exit 0; } [01:29] slangasek: ^^ that will mark the job as stopped [01:29] mathiaz: yes [01:29] slangasek: if euca_conf is not there [01:30] slangasek: however exit 1 will make the job fail if it's not configured [01:30] mathiaz: right - hence, asking what you want the behavior to be :) [01:30] slangasek: ok - just making sure I understand the semantics :) [01:33] slangasek: hmmm... so what's the difference between a failed job and a stopped job? [01:33] slangasek: the default configuration after package installation should be a working one [01:33] mathiaz: whether or not 'respawn' will try to start it again :) [01:33] (and what gets logged, etc) [01:33] I would expect eucalyptus-cloud to be a normal service, where you want init to be the process supervisor and respawn it if needed [01:34] slangasek: yes [01:34] so I think you want to change that 'exit 1' to a 'stop; exit 0' and then add 'respawn' [01:34] slangasek: even if there isn't any respawn option? [01:34] slangasek: ok [01:34] mathiaz: updated again, have a look there [01:34] slangasek: what would 'exit 1' + no respawn option do? [01:35] slangasek: in the case of a non-configured system [01:35] mathiaz: result in the job being marked failed, with upstart not trying to restart it [01:36] but, if you want upstart to *ever* restart your job for you when it dies, you need 'respawn', and then you want to write your job to avoid upstart getting into a busy loop [01:37] slangasek: busy loop meaning that the job would always fail and upstart would always restart it? [01:40] Upstart will stop trying if it’s respawning too fast for a while. [01:41] the difference between exit 1 and { stop ; exit 0; } is, basically, in the return code ;) [01:41] a job that just does exit 1 is marked failed [01:42] the event tells everybody that you failed [01:42] and that it was your pre-start that failed [01:42] the user who runs "start" gets told it failed, and gets a "1" return code [01:42] etc. [01:42] { stop; exit 0; } means you're not marked failed [01:42] you just stopped again [01:42] the event says you stopped normally [01:42] kees - i've pushed the change to bzr now. do you want me to subscribe u-m-s or will you handle this? [01:43] chrisccoulson: I can handle that, thanks! [01:43] the user gets told (basically) that the job they wanted started, stopped again [01:43] etc. [01:43] kees - thank you too:) [01:43] probably the biggest difference is that you don't get a console message ;) [01:44] Keybuk: so for the use case of: 'service not configured' what is the option to tell the user to first configuring the service before retrying to start? [01:45] mathiaz: there's no method to do that right now [01:45] Keybuk: failed job or start/stopped job? [01:45] Keybuk: which are the existing options for now [01:46] as slangasek says, we've settled on stop [01:46] Keybuk: ok [01:46] in the next version of Upstart, you won't need that at all [01:46] mathiaz: are you assuming that the output of the pre-start script will be passed to the user calling 'start'? [01:46] you can just ship your package in manual mode [01:47] (then it won't be automatically started - but will be started if the sysadmin starts it by hand) [01:47] Keybuk is planning to implement output logging for jobs. With it, it would be easy to print what the script said when e.g. the start command fails. [01:47] slangasek: hm not really. [01:47] mathiaz: ok good ;) [01:47] slangasek: I'm just asking questions about how upstart works :D [01:48] mathiaz: the problem is I tend to answer such questions with the question "how do you think it should work?" [01:49] ion: yes, but Keybuk is also planning to implement manual mode - which will do away with the need to say "aha! you didn't enable it" [01:49] Keybuk: ah ok. Than I'd like to be able to tell the user to configure the service by editing this file before attempting to restart the job [01:49] mathiaz: you can't do that [01:50] so ... how bad is it to do a Pre-Depends if you can't see a reasonable way without it? [01:50] Keybuk: right - I understand that. I'll use the stop+exit0 for now [01:50] personally I strongly dislike that kind of message anyway [01:50] "start foo" ... "but you didn't enable foo" [01:51] "yes, but I just told you to start it, stop being stupid and do what I told you" [01:52] Keybuk: the use case here isn't /enabling/ foo, it's /configuring/ foo [01:52] slangasek: well, that's different ;) [01:52] Keybuk: so you don't consider that upstart could help the sysadmin in diagnosing wrong configuration? [01:52] yes, it is [01:52] :) [01:52] Keybuk: a similar use case is to check for incompatible options [01:53] in that case, you probably *do* want to exit 1 [01:53] * Keybuk misunderstood, sorry [01:53] Keybuk: right - that's a good use of failed job [01:54] Keybuk: IIUC there isn't for now good logging? [01:54] Keybuk: "respawn" only affects the main script, not pre-start, yes? [01:54] slangasek: we don't respawn pre-start scripts ;) [01:54] * slangasek nods [01:54] Keybuk: like being able to say: your job has failed because foo option is incompitble with bar [01:54] but a pre-start script failed stops the job in a way that won't be respawned [01:55] who here knows how to grab translations out of rosetta for, e.g., ubiquity? [01:56] slangasek: seems like your eucalyptus job is doing what it's supposed to then [01:56] mathiaz: except Keybuk suggests 'exit 1' is better for the unconfigured case [01:56] instead of stop ; exit 0 [01:56] huh? [01:56] slangasek: mind if I upload this: http://launchpadlibrarian.net/33762653/moodle_1.9.4.dfsg-0ubuntu3.debdiff ? [01:56] if it's unconfigured, or the configuration is wrong, that's a failure [01:57] you want that communicated [01:57] it's not the same as "you didn't put splash on the kernel command-line" (which is a normal condition meaning the job is disabled) [01:57] Keybuk: ok - gotcha now. [01:59] LaserJock: what's the libc6 change that makes this necessary? [01:59] LaserJock: oh, it's doing IPv6 [02:00] slangasek: yep [02:00] slangasek: you get a 403 when trying to configure moodle without that patch [02:00] LaserJock: I think you probably want to list both values, then? Otherwise, you'll also be denying connections to on systems that have dynamic IPs, which is arguably a regression? [02:00] slangasek: ok, makes sense [02:00] (since /etc/hosts is set up with 127.0.1.1 , and users may expect that to work) [02:01] kees: has eucalyptus 1.6~bzr931-0ubuntu4 been already uploaded? [02:01] kees: the bzr branch still show UNRELEASED [02:01] slangasek: but good to go with that change? [02:01] damnit, fingers, remember how GRUB1 works [02:02] LaserJock: I think so, yes === lfaraone_ is now known as lfaraone [02:20] oh [02:20] that's neate [02:21] echo 0 > /sys/devices/virtual/graphics/fbcon/cursor_blink [02:27] slangasek: would you have any objection if we did something like [02:27] tty1 - boot messages, then a getty [02:27] tty2 - splash, then X [02:27] tty3+ - more gettys, if configuration allows [02:29] (lynx obviously) [02:30] Where do the other Xes go? [02:30] that's what I mean [02:31] ie. we hardcode something to own tty1 [02:31] we make usplash appear on tty2, then move it out of the way for X [02:31] X will claim the first free tty (taking out the tty7 hardcode), which will be tty2 [02:31] then tty3+ will be gettys and Xs depending on configuration [02:31] each taking the first free [02:32] if we don't boot with splash, you stay on tty1 until X chvts [02:32] and then you see useful messages [02:32] ah, neat [02:33] if boot appears to hang while usplash is up Ctrl-Alt-F1 will take you to useful messages [02:33] (Alt-F1 even) [02:33] Ctrl-Alt-F1 will still work in X [02:35] we could use tty7 for splash/X but then we need to keep the ugly hack that makes it go there [02:35] the hack that I keep breaking by accident :p [02:36] and then if you had only one getty configured, you'd have the weird thing where the second X login would be on tty2 ;P === TheMuso` is now known as TheMuso [02:48] HI! I was wondering if anyone has noticed the default netcat behavior for waiting on STDIN has been changed. It appears that it indefintely waits on stdin, whereas the default use to be to terminated immediately upon EOF [02:49] this seems to be a result of the addition of the -q flag, but I can't be entirely sure [02:52] docro: I don't think that's changed [02:52] netcat has always been a bit weird about EOF [02:53] I know at least in suse distros it termiantes immediately on stdin closing [02:53] echo "foo" | netcat server.com [02:53] will close once foo is written [02:53] let me grab the version #s real quickl [02:55] ubuntu: [v1.10-38] ... SLES 11: [v1.10] [02:55] SLES11 does not offer the -q option in its version of netcat [02:56] docro: suse could have patched it [02:56] Keybuk are you running ubuntu right now? [02:56] I doubt they would patch functionality OUT of netcat [02:56] docro: yes. [02:57] what version? if not 9.04 can you test the behavior [02:57] 9.10 [02:57] what version of netcat? netcat -h will show [02:57] as you say, v1.10-38 [02:58] hmmm I guess I would need to troll through commits to netcat in both distro to figure it out [02:58] the documentation says without -q, it will not quit on EOF on STDIN [02:58] for now I have juas aliased netcat to netcat -q 0 [02:58] Keybuk: the ubuntu documentation, yes [03:00] docro: just about every tool has minor differences between distros [03:00] Keybuk: agreed, it's just causing me grief :) I was wondering if ubuntu ever worked the other way [03:00] I don't think so, we inherit from Debian [03:01] not closing on EOF is not really a minor difference [03:01] according to the changelog, -q isn't even *in* the upstream netcat ;) [03:01] it looks like netcat is one of those "maintained by patches" tools [03:02] I have lots of code that depends on netcat closing upon EOF to gather results [03:02] because that is how it works in red hat and suse [03:02] it looks like there's netcat-bsd and netcat-traditional in Ubuntu as well [03:02] hmm [03:02] Maybe I will try those [03:02] thanks :) [03:02] docro: and just as many people (more in fact, we have more users ) have code that depends on netcat not closing on EOF [03:03] Keybuk: you have more desktop users maybe :P [03:03] enterprise is dominated with RH and SUSE [03:03] so? :) [03:03] who pays the bills :) [03:03] they don't pay our bills [03:04] not yet hehehe [03:05] Thanks for your help though, I will try the traditional package. Otherwise I will just leave my alias in [03:05] docro: think OEMs, ODMs, SoC chip/CPU designers (... companies that actually have *serious* amounts of money) [03:06] sladen i work for a company with serious amounts of money [03:06] not that my nagging would make them pay up! [03:06] docro: I'd approach Canonical for a support contract :) [03:07] sladen: currently one of the lone wolves moving towards desktop ubuntu for dev box and trying to get our tools working :) Doubtful our servers will ever go that way :( [03:08] alright good night, thanks for the help again [03:10] he was pleasant [03:10] normally there's more depramming of toys [03:11] heh, depramming of toys, I like that [03:39] kirkland: have you seen https://code.launchpad.net/~mathiaz/eucalyptus/karmic-fix-cloud-upstart-job/+merge/13452? [03:39] kirkland: for https://bugs.launchpad.net/ubuntu/+source/eucalyptus/+bug/452665 [03:39] Launchpad bug 452665 in eucalyptus "eucalyptus-cloud runs without any option set" [High,In progress] [03:39] mathiaz: the upstart fixes? [03:39] mathiaz: yeah [03:39] mathiaz: why the respawn? [03:41] kirkland: if eucalytpus-cloud dies, you wanna get it restarted no? [03:41] mathiaz: i think so; just didn't see a note about that in the changelog [03:43] mathiaz: ifconfig eth0:1 169.254.169.254 netmask 255.255.255.255 [03:43] mathiaz: that looks to be working for me [03:44] kirkland: you may wanna call it differently [03:44] kirkland: like eth0:eucalocal [03:44] mathiaz: i couldn't get it to work with a string [03:44] mathiaz: only with an int [03:44] kirkland: avahi creates an eth0.avahi [03:44] kirkland: IIRC [03:45] kirkland: hm - may be you don't wanna have a virtual interface [03:45] mathiaz: i couldn't get it to work with the ip command either [03:45] mathiaz: had to use ifconfig [03:46] kirkland: ip ... label xyz [03:48] sladen: hmmm [03:48] sladen: sudo ip addr add 169.254.169.254/32 scope link dev eth0 label eth0:1 ? [03:49] ... dev eth0:myfoo label eth0:myfoo [03:50] ScottK: ping [03:50] sladen: ubuntu@cluster:~$ sudo ip addr add 169.254.169.254/32 scope link dev eth0:9 label eth0:9 [03:50] RTNETLINK answers: File exists [03:50] kirkland: [03:50] j1mc: [03:51] kirkland: have you removed the existing 169.254.169.254/32 ? [03:51] kirkland: try adding an address that isn't in-use, eg .169.253 [03:51] sladen: mathiaz; bingo [03:54] mathiaz: would you like me to merge and upload your upstart changes? [03:54] mathiaz: i have no objections [03:54] kirkland: yes please [03:55] Keybuk: I would object to us changing that in karmic ... :) [04:06] mathiaz: how much longer are you online [04:06] kirkland: don't know [04:06] kirkland: why? [04:07] kirkland: are you off tomorrow? [04:07] mathiaz: i was thinking about it, but i didn't hear from alice [04:07] mathiaz: i wanted you to review my eth device labeling [04:07] kirkland: why - do you have a fix to test? [04:08] mathiaz: i'm testing it now [04:08] kirkland: sure - send a branch [04:08] mathiaz: k [04:09] mathiaz: the lintian output for eucalyptus is a mess [04:16] mathiaz: brilliant! [04:16] sladen: thanks for the pointer! [04:16] kirkland: I know - thanks :) [04:28] mathiaz: lp:~kirkland/eucalyptus/eth-label [04:35] mathiaz: looks like it's working to me [04:36] kirkland: I'll build a local version here and run through my tests [04:38] Hi there. My package needs updating from Debian testing. Is this the right place? [04:51] kirkland: 1, scorpion: 0 [04:52] http://rookery.canonical.com/~kirkland/CIMG0097.jpg [04:52] back to eucalyptus [04:53] kirkland: hm - build your branch [04:53] kirkland: upgraded packages - still 169.X on eth0 [04:53] mathiaz: i had to reboot [04:54] mathiaz: restarting networking and eucalyptus might do it, though [04:54] kirkland: does this mean that on shutdown eucalyptus-cloud doesn't remove its ip address? [04:54] mathiaz: probably [04:54] kirkland: well - the package upgrade should restart eucalyptus-cloud [04:55] kirkland: well - I've just rebooted my CC [04:56] mathiaz: i'm fixing it to handle the pub and priv addresses too, with a label [04:56] mathiaz: eth0:pub and eth0:priv [04:56] mathiaz: for the guest ip's [04:56] mathiaz: which are also stacked on eth0 [04:56] kirkland: hm [04:56] kirkland: are you using multiple addresses? [04:56] mathiaz: ? [04:57] kirkland: there can be more than one public IP assigned to the CC [04:57] kirkland: like .2 .3 .4 .5 [04:57] kirkland: all assigned to the CC [04:57] mathiaz: i'm not using an enumerator right now [04:57] mathiaz: that's more complex [04:57] mathiaz: mainly, i'm getting the superflous addresses off of eth0 [04:57] kirkland: right - just saying - multiple public IPs are common [04:57] mathiaz: and onto their own label [04:58] kirkland: right - so eth0:pub would have multiple public IPS [04:58] mathiaz: i'm just fixing the ones that eucalyptus creates for vm guests [04:58] mathiaz: right [04:58] kirkland: ok - if that works well that's great [04:59] mathiaz: i'm pushing an update now [04:59] mathiaz: did your CC come back up? [04:59] kirkland: yop [04:59] mathiaz: actually, pushing to lp:~kirkland/eucalyptus/eth-label2 [04:59] mathiaz: and? ifconfig says? and ip addr says? [05:00] kirkland: the eht0:metadata is there :) [05:00] mathiaz: nice [05:00] kirkland: let me check if instances are working correclty [05:00] mathiaz: please [05:00] kirkland: http://paste.ubuntu.com/294367/ [05:01] mathiaz: looks good [05:01] mathiaz: and ifconfig? [05:01] kirkland: well - there is still 169.X in there [05:01] kirkland: ifconfig properly shows eth0 and eth0:metadata [05:01] mathiaz: pastebin that [05:02] kirkland: http://paste.ubuntu.com/294368/ [05:02] mathiaz: good [05:02] mathiaz: inet addr:192.168.12.118 Bcast:192.168.12.255 Mask:255.255.255.0 [05:02] mathiaz: that's the key one [05:02] kirkland: yes [05:02] mathiaz: i'm please with that [05:03] * mathiaz nods [05:03] mathiaz: okay, branch lp:~kirkland/eucalyptus/eth-label2 [05:03] mathiaz: sorry, had to push a new branch [05:03] mathiaz: i uncommitted something, and things diverged [05:03] kirkland: now eucalyptus-ipaddr.conf uses ip addr [05:03] kirkland: ok [05:03] mathiaz: yeah, i think it should use ifconfig [05:03] mathiaz: but that's just my preference [05:03] mathiaz: i'm not going to change it, but i would support someone else in doing so [05:03] kirkland: so I was poking around the configuration and found something interesting in /etc/eucalyptus/eucalyptus.conf [05:04] mathiaz: ? [05:04] kirkland: VNET_PUBINTERFACE="eth0" [05:04] VNET_PRIVINTERFACE="eth0" [05:04] kirkland: seems that we already know which interface should be used for private /public interface [05:04] mathiaz: i suspect that's intended for servers with multiple eth cards [05:04]