[00:00] and other packaging bits [00:00] I'll try again [00:01] what upstart file? [00:01] http://paste.ubuntu.com/1228509/ [00:01] you're talking about that change? [00:02] the file is not being written at all [00:02] that change there would have no affect on it being written. [00:02] please wait [00:02] (and just random bit of information, [00:02] sh -c 'echo -n foo > /tmp/bar; read a < /tmp/bar; echo $?;' [00:02] ) [00:03] that does return '1'. [00:03] if you dont have a trailing new line [00:03] its annoying. [00:03] that would explain a lot [00:04] sigh, it's now impossible to use mouse select on konsole when byobu is running [00:04] its dirty magic, that [00:04] somehow tmux actually has "mouse support" [00:04] i was confused by this once too. [00:05] its doing something magic with your mouse [00:05] it seems to be when the clock ticks over each second [00:05] the selection is reset [00:05] ah. [00:05] well, byobu-config [00:05] and you can turn the clock off [00:05] yeah [00:05] or highlight faster [00:06] I need more bandwidth [00:06] interesting. [00:06] (use canonistack) [00:06] interesting... [00:06] can't do that with my test hardware here [00:06] i'm seeing the "No boot images have beeen registered yet" error on maas again. [00:06] I suspect because the celery stuff is not working [00:07] (although i successfully booted off of said unregistered image and enlisted [00:07] ) [00:07] there were some major routing changes [00:07] and there was fallot [00:07] fallout [00:08] bigjools, i suppose you're lamenting the ephemeral images [00:08] when you mention bandwidth [00:08] right? [00:08] no, quantal updates [00:08] ah. [00:08] well, can't really hep you there. [00:08] I already run a caching proxy [00:08] but i was just thinking i should write mirro utility [00:08] which doesn't help on the first run :) [00:09] (the maas.ubuntu.com/ only comes over https, so squid sucks) [00:09] urgh [00:09] and.... pulseaudio... ffs [00:09] but it'd be pretty easy to right a mirror of it. [00:09] well, you probably can't report a bug on it because glib updated yesterday and your out of date. [00:10] :) [00:11] ok I am going to spin up a canonistack and install from scratch [00:11] bigjools, just as another variable to play with, there are now daily images of both quantal and precise that "should work" (they've tested as working for me). [00:11] i'm sure you've seen that script i use [00:11] i dont run it as a script [00:11] but i just copy and paste bits together [00:12] hey, i'm just laying. [00:12] i just registered a node [00:12] commissioned it [00:12] now on its node page i see "start node" grayed out [00:13] ah. needed to add a key. [00:13] which quantal image is best? [00:14] just use latest I guess [00:14] bigjools: Do you happen to know what else is needed to save a setting? http://paste.ubuntu.com/1229537/ [00:14] is there 2? [00:14] daily is what i meant. [00:15] roaksoax: I think that should work [00:15] bigjools: it doesn't :( [00:15] your label should say series not release :) [00:15] alright. i'm out for the night. [00:15] cheers smoser [00:16] roaksoax: ummm [00:27] roaksoax: I'm not sure, sorry. Django is full of traps and I hate it. I'd need to ask rvb [00:29] bigjools: yeah i was planning to ask him [00:30] gah byobu [00:30] can't mouse select even with its clock off [00:31] roaksoax: I can't install maas on a clean canonistack instance, it's blowing up at the db config stage saying there's no pg clusters [00:32] and when the debconf message appears it goes over the whole screen obscuring the error underneath >:( [00:35] bigjools: pastebin? [00:35] roaksoax: any ideas? http://paste.ubuntu.com/1229558 [00:36] bigjools: is postgresql installed and running? [00:36] bigjools, i told you. its magic with your mouse. [00:36] roaksoax: installed but not running === matsubara is now known as matsubara-afk [00:36] smoser: it is :/ [00:36] http://awhan.wordpress.com/2012/04/18/tmux-copy-paste-with-mouse/ [00:36] the solution is to use the shift key. [00:36] thanks [00:37] (note, i wasn't just holding out, i didn't remember. i think i asked kirkland once) [00:37] smoser: not working doing that either [00:37] hm.. [00:44] roaksoax: so installing postgresql on its own doesn't set up a cluster and it's not running afterwards [00:51] bigjools: to me it looks that postgresql is not running [00:51] roaksoax: it's not - I had to manually create a cluster [00:51] then everything works [00:52] bigjools: ah! so maybe there's an issue with postgrsql package? [00:52] possibly [00:52] easy to re-create [00:52] * bigjools debugs non-working dhcp [00:53] bigjools: it doesn't look like postgresql actually. The last package was uploaded long time ago: https://launchpad.net/ubuntu/+source/postgresql-9.1/9.1.5-2 [00:54] dbconfig-common hasn't been updated either [00:54] something odd is going in [00:54] on [00:54] so there's some other reason [00:54] maybe fire up another instance, update, upgrade [00:54] and try again? [00:54] I just started a new quantal instance, sudo apt-get install maas [00:54] and it does it [00:55] roaksoax: there's another bug in that installing maas-dhcp doesn't terminate the existing dhcp [00:55] or prevent it running at all [00:56] bigjools: yeah I saw the report but haven't been able to take care of any packaging related activities today [00:56] bigjools: btw.. we need a way to disbale dhcp from the command line, i/e maas-provision disbale-dhcp [00:57] roaksoax: ok [00:57] so that when maas-dhcp is recmoved, or reconfigured to not have it enabled, it does not stay up and running [00:57] roaksoax: we're going to change it so the maas-dhcp package doesn't configure dhcp at all, just installs packages [00:57] it's really screwing up things with the way we want to configure dhcp otherwise [00:58] bigjools: that's fine too [00:58] it would makes things simpler [00:58] completely [03:13] bigjools: i also saw that issue once [03:13] didn't see it again though [03:13] roaksoax: the rabbit one? [03:13] bigjools: yeah [03:17] hi jam [03:17] hi jtv I mean [03:17] Hi bigjools [03:17] Being greeted is so much better as a way to start your day than a corrupted filesystem. [03:17] What a shame the corrupted filesystem was here first! [03:17] bugger [03:18] Well, it's fixed now. But I used to get this as a recurring problem, so it doesn't bode well. [03:21] Problem with the nvidia driver we suspect. [03:24] roaksoax: still there? [03:24] jtv: damn [03:24] all ok now? [03:24] For now, yes. [03:26] roaksoax: in my maas-cluster-controller.postinst I am doing a db_get to get hold of the url to maas but only on configure/reconfig at the moment, but this means that in normal install/upgrade circumstances it doesn't ask for it and breaks the end config [03:26] * roaksoax is her unfortunately [03:26] I am hoping that I can ask for it in installation as well without violating any rules [03:26] roaksoax: :) [03:27] bigjools: uhmmmm [03:27] jeeepers, Australia has bug butterflies [03:27] big* [03:27] * roaksoax is doing homework again :) [03:28] let me see [03:29] bigjools: right, so there's something wrong there [03:29] roaksoax: what are the rules on using db_get? [03:30] bigjools: let me look for an example [03:30] can't ask during upgrades IIRC [03:30] bigjools: remember, it is not an upgrade [03:30] bigjools: it is a new install [03:30] bigjools: so it *will* have to ask [03:30] roaksoax: that's not true [03:30] the new maas pulls it in [03:31] seed cloud case [03:31] bigjools: so if it pulls it in, what are you trying to achieve with a db_get ? [03:31] roaksoax: in other circumstances it is a stand-alone app running somewhere else [03:31] it has to ask [03:32] but in the seed-cloud case I am wondering if we can just pull the URL out of maas's dbconfig [03:32] bigjools: hold on hold on [03:32] * bigjools holds [03:33] bigjools: so 1. i install maas-cluster-controller and I should receibe a question asking me to input the maas-url ? (this maas-url is the region controller?) [03:33] roaksoax: yes [03:33] and yes [03:33] bigjools: ok, so what's seed cloud again? [03:33] roaksoax: the original 12.04 package [03:34] maas becomes a virtual package and pulls in region & cluster [03:34] bigjools: right, so the workflow is that every time maas-cluster-controller is installed for the first time, it should request the admin the URL right? [03:34] roaksoax: yup [03:35] roaksoax: also [03:35] bigjools: ok so an upgrade from previous maas is the same case [03:35] it needs to ascertain it on upgrade from old maas [03:36] bigjools: right, so we can grab that from celeryconfig [03:36] which is maas/default-maas-url [03:36] bigjools: rigth, so hold on [03:36] * bigjools holds on [03:36] bigjools: normally.... eveyr time maas-cluster-controller is installed for the first time it should ask for the URL [03:36] every time except when upgrading from the 12.04 maas :) [03:37] it doesn't matter if we are upgrading from 'maas' to maas-{region,cluster}-controller [03:37] oh? [03:37] the package is installed for the first time [03:38] so it *should* ask for it, don't you think? [03:38] so when someone with maas installed on 12.04 upgrades to 12.10, I thought it was bad to ask questions [03:38] i don't know to what extend it is bad... but TBH.. we are not free from not asking them [03:39] I think we are though, because it will be in maas/default-maas-url [03:39] bigjools: i mean... I recall having had to answer questions of this sort before [03:39] bigjools you could ask debconf db for the value, sure [03:39] so if I can work out that we're upgrading from old maas... [03:39] I don;t know how to work that out [03:39] bigjools: we could, I don't know whether is a sane thing to do though [03:40] in the way of policy compliant [03:40] as in quering debconf db for another's package db [03:40] ok [03:40] perhaps grep celeryconfig then [03:40] bigjools: that's not all [03:41] bigjools: consider that if we still all-in-one MAAS server from the CD installer, we should not ask that question [03:41] * bigjools 's head explodes [03:42] bigjools: so looking at celery might be agood idea, but what happens if cluster-controller gets installed *before* region-controller [03:42] if celeryconfig is not there, we're not upgrading so we have to ask [03:42] bigjools: right, but what if we don't know it? [03:43] bigjools: as in the administrator doens't know it [03:43] there's the $64m question [03:43] bigjools: i'll have to think about that though [03:44] bigjools: but anyways, the db_get and stuff [03:44] bigjools: see at maas-dhcp.config that will show you how the question is asked [03:44] bigjools: basically, the .config should ask the questions [03:44] db_input high maas-dhcp/maas-dhcp-range || true db_go [03:44] db_input high maas-dhcp/maas-dhcp-range || true db_go [03:44] err [03:44] db_input etc etc [03:44] db_go [03:45] and in .postinst [03:45] db_get should be used to obtain the value [03:45] I already do this [03:45] but only on configure [03:45] or reconfigure [03:45] it doesn't seem to get called on install [03:46] oh [03:46] bigjools: so look at region-controller.postinst for what I do in reconfigure [03:46] I am missing a .config! [03:46] to set the new value [03:47] yeah [03:48] roaksoax: I still think this is a problem [03:49] for dist upgrades [03:49] indeed [03:49] we'll have to get them tested very well [03:49] we not only have to consider dist-upgrade,s but also CD installs [03:49] in maas.config it avoids asking questions on install [03:50] bigjools: that's a hack for dbconfig-common [03:50] I don't know how to detect all these conditions [03:50] bigjools: bigjools and to not ask for default-maas-url, becuase we were "attempting" to determine it automatically [03:51] so maas.postinst was a special case because they didn't want dbconfig-common questions being shown on install [03:51] nor ask for the IP MAAS used [03:51] so we made a first attempt to determine it [03:51] that is why often times [03:51] we have to run a reconfigure [03:51] ok [03:51] to set the IP of the internal network or network that is connected to the clients [03:52] bigjools: we cannot just do that on cluster-controller because is a completely different case [03:52] indeed [03:52] unfortunately :( [03:52] look, I'll set it up now so it asks questions so I am unblocked, we can fix it later [03:53] It's almost in a usable state nw [03:53] now [03:53] awesome! [03:53] i'll play with it tomorrow [03:53] well it installed [03:53] that's a start :) [03:53] I just have this config proboem [03:53] bigjools: hehe, well you got the easy part [03:53] :) [03:53] :) [03:53] separating is much easier than packaging from scratch :) [03:53] I've never made a package before [03:54] i suffered hard when i first packaged maas :) [03:54] yeah well I cobbled up the first revision of packaging too, because I just used a tool to do it :) [03:54] heheh [03:54] well that's how you learn [03:54] whose name I forget [03:54] hands on experience [03:54] totally [03:54] anyways [03:54] i'm off [03:55] good night, and thanks roaksoax [03:55] have a good rest of the day [03:55] yeah, lunchtime is here [04:34] what does maas do if an api user asks for a non precise/quantal series? [04:38] jtv: why is the nvidia driver corrupting your filesystem ? [04:40] jam: I don't know if it is. But the last time I had regular filesystem corruption on this system, William had the same problem at the same time using the same driver. Plus we know the problems with nvidia... [04:41] roaksoax, fwiw merged the branch, and fixed the broken test [04:43] jtv: do you suffer machine lockups ? [04:43] huwshimi: I'd like to chat with you quickly about the UI changes for hardware constraints. We've done most of the backend stuff, but I'm not 100% sure where to put the frontend changes. [04:44] jam: Sure [04:45] lifeless: no, no lockups. I suspect memory corruption just hitting the filesystem because the filesystem keeps so much data in memory. [04:45] Lots of crashes, of course, and no unity. [04:45] But that may be normal unrelated beta fare. [04:45] machine crashes ? [04:45] spontaneous reboots/OOPSes ? [04:47] jam: Hangout/irc? [04:48] huwshimi: well, IRC is fine for me, but we could do hangout, too. [04:48] jam: IRC is fine. [04:49] huwshimi: so I think the basic idea is to expose: cpu_count/mem/arch/tags to the Node page itself [04:49] at least as a first step [04:50] jam: Yup. [04:50] huwshimi: so what do I actually edit on my end to make sure that stuff gets exposed for you. [04:51] is it updating the 'views' or the 'forms' or the HTML template code? (All of the above?) [04:52] hazmat cool thanks [04:52] allenap: around already¿ [04:52] jam: You'll need to expose the hardware info in NodeView (src/maasserver/views/nodes.py) [04:53] (I think that's the right place) [04:53] jam: I can then expose them on the html side [04:54] huwshimi: k [04:55] jam: I'm not sure about the specific implementation, but it will be something along the lines of adding the data to the context for rendering (I'm not sure how familiar you are with Django). [04:56] huwshimi: not very, but I'm pretty good at 'google: django XXX' :) [04:56] hehe [04:56] jam: You can test it by outputting the same variable names somewhere in the template (with {{ var_name }}) [04:57] roaksoax: allenap only went to bed an hour ago :) [04:58] heh :) [04:58] roaksoax: got the dbconf working for now [04:58] jam: Also, somewhere there will be a place to test the view is returning the correct data. [04:58] needs fixing for upgrade + CD install cases but it's a starty [04:58] awesome [04:58] indeed [04:58] alright illntest tomorrow [04:58] typos R us today [04:59] jam: Ah, src/maasserver/tests/test_views_nodes.py and you can test the client response contains the vars [05:00] huwshimi: thanks [05:01] lifeless: not experiencing reboots, no. Individual programs crash. [05:03] jtv: ah, how annoying [05:03] jtv: anything in dmesg ? [05:03] What sort of thing are you thinking of? [05:04] (I also had to switch to KDE for the duration, so a poor comparison base) [05:04] you sometimes get warnings on bad memory reads [05:04] Oh [05:04] or crashes in kernel threads [05:04] The last time I had the FS corruption problem I did a full memory check. Showed no problems. [05:05] Nothing in dmesg that jumps out at me. [05:06] I'll have a closer look on my next few crashes. [05:10] bigjools: how about I remove the uuid option from start-cluster-controller but add the uuid generation/loading in a separate branch? [05:10] jtv: sounds good [05:10] ok [05:12] sorry I missed your reply to the MP, I'm not checking email religiously ATM [05:13] jtv: python-netifaces!!!! [05:13] Yeah [05:14] jtv: we should check with rbasak that it works on arm [05:53] bigjools: mgz and I won't make it to the standup today. He is on a train going to London, and I have to go pick up my son from school (it is an early-out day). [05:53] ok thanks for letting me know [05:54] We are essentially polishing the Tag api a bit, I'm looking to expose the new fields for the HTML, and working up a list of things for Diogo to test. [05:54] I'm also personally running into 'make run' doesn't work on my Quantal VM [05:54] Diogo is going to be a busy man [05:54] what's happening? [05:54] Well, a lot of our testing can be automated, I think. Which is meant to be done as part of the Jenkins w/ packages, right? [05:55] bigjools: well, something starts up, but then I get proxy errors if I try to connect to the web ui. [05:55] I'm also wondering if it is related to my other rabbit failures. [05:55] I can connect to port 5240, but 5243 (?) is failing to respond [05:56] I've had this before [05:56] 'bin/test.maas' is failing a Rabbit test, and hanging on another one. [05:57] and 'make run' doesn't seem to be bringing up a service, but it is hard to sort through the chatter to figure out what is actually wrong. [05:57] jam: manually starting the regular rabbit service didn't work? [05:57] jtv: it is already running [05:57] restarting it didn't help [05:57] :( [05:58] Just now I did uninstall and re-install the rabbitmq-server, but it still failed in the test suite. [05:58] jam: did you try ctrl-C'ing the test to see the traceback? [05:59] jtv: I can get that, (which is how I know it is a rabbit test) [05:59] I can paste it for you. [05:59] In 'ps' I do see a beam.smp process running as a child of the test runner. [06:00] jtv: http://paste.ubuntu.com/1229812/ [06:02] 5243 is longpoll isn;t it? [06:03] jtv, bigjools: It looks like I'm getting a 'logs/webapp/current' message which is lockfile.LockTimeout [06:03] So maybe there is a stale lockfile somewhere? [06:03] where would that be? [06:03] in /run [06:03] Oh, the /run/maas.* [06:03] jtv: no files named ./run/maas [06:03] only named exists there [06:04] no . [06:04] / [06:04] /run/maas.* [06:04] They have a tendency to stick around, don't they? [06:04] database one sometimes does for me [06:04] bigjools: /run is owned by root, and there is no '/run/maas' directory [06:04] are we supposed to be doing 'sudo make run' ? [06:04] no [06:05] hmm [06:05] what's going on [06:05] Not /run/maas/, as I recall, but /run/maas.* [06:06] jtv: ls /run/ma* no such file or directory [06:06] /run/lock maybe? [06:06] That's a directory. [06:06] jtv: lots of /run/lock/maas* files [06:06] just nuke those? [06:06] ah that's it :) [06:06] Better make sure their associated processes are dead. [06:07] jtv: no python processes running that are 'maas' related. [06:08] Doesn't have to be python IIRC. [06:08] The files contain pids, don't they? [06:08] however, I still get 'Service Temporarily Unavailable' on localhost 5240 [06:08] jtv: have we got a quick way of seeing if a string matches an enum value? [06:08] bigjools: map_enum [06:08] jtv: well those files don't exist anymore now... :) [06:08] You turn the enum into a dict, then see if the string is in there. [06:08] jtv: and looking at it now that they exist again, they are all 0 length files. [06:08] no pids [06:08] Erm. Haven't had these problems for a while so don't recall which processes they were. :( [06:09] Try ps -ef | grep maas [06:09] (The "maas" is often in surprising places IIRC, such as the path for the executable; you're not looking for executables called maas) [06:10] jtv: only postgres and my editor [06:11] See any named running? [06:11] jtv: no, because the install recommended disabling named as part of installing bind [06:12] (I can run it in my vm, though) [06:14] Just checking -- it's one of those processes I found myself killing a lot when I had lock-file problems. [06:14] jtv: now I have named running, startup still seems to be not working properly: http://paste.ubuntu.com/1229823/ [06:15] Note that the lock file that seems to be going stale is "maasserver.start_up.lock' [06:15] I see that there *is* such a process [06:15] Maybe it's not stale, just blocked... What were we supposed to be running on 5423? [06:16] (because that file has the same inode as link-vm.MainThread-29249) [06:16] which is an existing pid [06:17] Hmm... the API I guess. I thought that'd be on 5240. [06:18] jam: I guess apache could be implicated then. [06:19] jtv: ./logs/webapp/current still has a timeout trying to grab the FileLock. However, I see that the file exists on disk, and if I delete it, I still get a failure. [06:20] Sounds as if the lingering lock is a symptom, not the cause. [06:20] In other words, maybe it's not stale, just blocked on something. [06:22] jtv: my immediate thought is that the timeout passed to LockTimeout is being misinterpreted. The docs say it is 'seconds', but what if it is 'milliseconds' [06:23] hmmm... the code looks to be using time.timeout, which seems like it would be fine. [06:23]