[01:00] if im gonna use maas for virtual machines and juju i have to install LXC on my MAAS server correct? Not to familiar with this yet, but running lxc-create now for node, hope im on the right track === CyberJacob|Away is now known as CyberJacob [08:18] Hi jtv, time for medium-sized review? https://code.launchpad.net/~rvb/maas/cluster-own-ui4/+merge/213848 [08:57] gmb: arg, found a critical bug in maas-test (https://bugs.launchpad.net/maas-test/+bug/1301815). [08:57] Ubuntu bug 1301815 in maas-test "maas-test is incompatible with the most recent MAAS packages (revision 2182 and up)." [Critical,Triaged] [08:58] gmb: I thought maas-test would still work (but drop the simplestreams config) but since it explicitly runs the (now gone) maas-import-ephemerals it breaks completely with recent packages. [08:58] rvba: Ack. So the fix is just to drop that call? [08:58] gmb: if /etc/maas/bootimages.yaml is there: drop that call [08:59] gmb: next step would be to add a way to customize /etc/maas/bootimages.yaml but that can be done later. [08:59] rvba: Right. Want me to take this? Once I’ve done reviewing I’m free. [08:59] gmb: Okay, thanks. [08:59] Cool. === CyberJacob is now known as CyberJacob|Away === wgrant_ is now known as wgrant [13:20] rvba: allenap: I was out yesterday. But the new import script does provide the UEFI files, and I have tested it with daily-qa, two days ago [13:27] blake_r: all right, thanks a lot. [13:41] Anybody want to review a validator for Node.hostname? https://code.launchpad.net/~jtv/maas/bug-1300059/+merge/214032 [14:18] jtv: what happens if someone has existing nodes with invalid hostnames? I mean, after this lands on existing installations with invalid hostnames. [14:21] They can't save with the broken hostname. [14:21] But they can keep it in the database for as long as they like. [14:23] jtv: which means you won't be able to change *any* field on these nodes. [14:23] Unless you also change the hostname. [14:23] Right. [14:24] I know that's harsh, but I have no sympathy at all. :) [14:25] jtv: I can see that :) [14:26] I agree with you on the principle… just trying to think what kind of pain this can inflict on users. [14:28] I think it's pain we simply have to go through. We could smooth it by warning about bad hostnames. [14:30] jtv: not sure it's worth it. It won't help people using the API/CLI at all. [14:31] Right... And to be honest, I hope that nobody will have gotten this wrong. Except maybe the underscores. [14:33] jtv: not entirely happy with the multiple error messages that a user could get when trying to enter invalid hostnames: [14:33] I can imagine a user spending quite some time trying to figure out what the allowed format is. [14:34] A network administrator not being able to figure out the allowed format for hostnames? I doubt it. [14:35] jtv: but I mean, the error message should not only tell you about the error, it should also state what a valid hostname should look like (even if it's just a mention of the proper RFC) [14:36] Can you give me an example of how you have gotten this wrong recently, and what documentation you looked up to resolve it? [14:36] Or preferably, somebody who manages systems and networks? [14:36] It just seems outlandish to me. [14:37] hey, how often does MAAS update it's ephemerals and PXE files? I'm using it to do a lot of trusty type dev and testing, and want to be sure that it's pulling whatever the latest trusty bits are. [14:37] It's either weekly, or manually only — let me check. [14:38] bladernr_: once a week. [14:38] jtv: yeah, I might be splitting hairs a bit. I'll review your branch now :) [14:38] I was under the impression MAAS was supposed to update itself once you did the initial update... if that's not the case, we definitely need to know as we're requiring MAAS for cert now :/ [14:39] rvba: Thanks. Maybe it's just the voice of laziness, but I feel "hostname" should be obvious to someone who manages hosts. :) [14:39] btw, branch up for review: https://code.launchpad.net/~rvb/maas/reporting-bug-1301809/+merge/213984 [14:39] I'll take it next. [14:39] Or you can leave it to gmb or allenap. :) [14:40] bladernr_: it refreshes its copies of the boot images once a week. [14:40] ahhh ok [14:40] so running maas-import-pxe-files is essentially a manual update that can be run whenever necessary, yeah? [14:40] Technically, that has nothing to do with that initial manual download. We just ask people to do the manual download so they don't have to wait a week to start working. :) [14:41] As you say, it's just a manual update that you can run anytime, in addition to the weekly update. [14:41] ahhh, ok. Hah... I thought there was some sort of magic sauce in MAAS that was triggered by the first download... tricksie you are [14:41] It can also be triggered from the UI. [14:41] jtv: btw, I'm quite surprise that we don't have a hostname validator somewhere that we can re-use… [14:41] surprised* even [14:41] So am I. The Apache foundation seems to have something in Java, but... [14:41] jtv: cool. Thanks for the confirmation/clarification. That's exactly what I needed to know. [14:42] np [14:42] rvba: Django has built-in URL and email validators, though I'm not sure they're as thorough as mine. [14:42] Actually, without meaning to brag, I'm fairly confident they aren't. [14:43] I considered tricks like "put http:// in front and see if it's a valid URL" — but that's susceptible to all sorts of mistakes. [14:43] rvba: ah _that_ MP... I saw that earlier and didn't review it because there was no commit message and no description. I thought it was a WIP. [14:44] jtv: I just came across http://python-jsonschema.readthedocs.org/en/latest/validate/#validating-formats [14:45] jtv: rarg, the commit message didn't get through. Fixed now. [14:45] Hadn't seen that one. I'll have a closer look. [14:47] rvba: the jsonschema one isn't very thorough either. :/ [14:47] It'll even disallow some names that I _think_ are valid. [14:48] On the other hand, it looks like it'll allow "..." as a hostname. :-) [14:48] jtv: how can you tell? It's not very clear in the doc what it does. Are you looking at the code? [15:20] rvba: yeah, I read the code. [15:24] jtv: question for you. I'd like to add a tiny notification next to 'Clusters' at the top of the page when some of the clusters need attention from the user. More precisely, I'd like to display the number of pending clusters + the number of accepted clusters with no boot images. [15:24] jtv: but I'm afraid this would be a bit too expensive to compute to display it on every page… [15:26] jtv: I didn't try to measure the time it takes to compute the number of accepted clusters with no boot images yet… but I guess I could try to see how it goes. [15:26] jtv: well, it's not really a question… but I'd welcome your opinion on that :) [15:27] I think it's a good idea, and I don't expect that much performance overhead. Clusters are smaller in the database than nodes, and we expect fewer of them. [15:27] The real question is what feedback to show. That sounds like it needs UI design. [15:27] The flavour-of-the-week UI, I guess, would be to show a red field over the Clusters tab with the number of items that need attention. [15:28] jtv: I've got an idea about that. [15:28] ? [15:28] jtv: exactly. I already have a couple of CSS class to do precisely that. [15:28] classes* [15:29] But much more a question for UX people than for Unix people, if you'll pardon the pun. [15:31] Right, I'll put a mockup together. === roadmr is now known as roadmr_afk === roadmr_afk is now known as roadmr [16:41] MAAS networking | http://askubuntu.com/q/442957 [17:01] gmb: did you do something earlier today to fix that 1.5 lander? My branch got landed but now I've got another one stuck… maybe I'm just not patient enough… [17:34] gmb: my bad, it got merged eventually. Just took forever. [17:43] rvba: is it possible to add 14.04 yet to MAAS to deploy servers with it? [17:45] blahRus: yes, but you it will be much simpler to do if you use a Trusty server to run MAAS. [17:48] rvba: we were getting internal servers errors all over the place with beta 2 [17:48] let me try it again === lazyPower is now known as lazyPower-travel [17:48] I'll try off of a beta [17:48] daily* [17:48] blahRus: a new version has been published recently. I've got reports that it works okay. [17:49] kk [17:49] 1.5+bzr2204-0ubuntu1 <- that's the version in Trusty now. [17:49] kk === CyberJacob|Away is now known as CyberJacob [18:40] rvba: I didn’t actually do anything; I tore the instance down and started a new one. The annoying thing is that I’ve no way of debugging the lander… it gets deployed with some random SSH key that I don’t seem to have access to (maybe bigjools does). [18:40] Anyway, yay for things being fixed just by wishing. [18:40] Next up, I shall wish for plug-and-go MAAS and see what happens. === roadmr is now known as roadmr_afk === roadmr_afk is now known as roadmr === sputnik1_ is now known as sputnik13net [20:54] hi allenap, smoser: can you tell me how the new import script works? [20:54] so when i import images [20:54] immediately after that is done i try and create a node and it fails [20:55] (tells me "there are no images") [20:55] but if i wait a bit, it works fine [20:55] presumably now it does more than a stat() on the images to see if they're there, my guess is a celery notification [20:55] what should i wait for before i can add new nodes? [20:56] tych0: Did you run maas-import-pxe-files directly? [20:56] yes [20:56] allenap: is there a better way to trigger it so that it will automatically get updated? [20:56] It can take a few minutes after its finished before the region susses out what images are available. [20:56] yep, understood [20:56] how do i know? [20:56] maas boot-images report-boot-images or something? [20:56] tych0: You can run it from the UI. [20:57] not in the cloud installer i can't [20:57] Ah :) [20:57] me looks [20:58] tych0: bin/maas foo node-groups import-boot-images [20:58] allenap: that won't block until i'm done, though [20:58] so i still need a way to check [20:58] tych0: bin/maas foo boot-images read $cluster_uuid [21:00] allenap: cool, that'll play, thanks [21:00] tych0: So, there’s a check every 5 (or 10?) minutes that’ll find the imported images. Importing images via the API (or UI) ensures that the images are reported as soon as the import run finishes. [21:01] ok [21:01] allenap: is there any way to tell it to look? [21:01] and/or tell the m-i-e script via the API to use an http proxy? [21:01] tych0: You /might/ find that the periodic check picks up a half-finished import, but I suspect it won’t; the new import script switches a symlink when it finishes. [21:02] actually i think it should be ok [21:02] we're not even pxe booting the first node [21:02] so we don't even really care if the images are even there [21:02] tych0: maas-import-ephemerals is an ex-script. maas-import-pxe-files isn’t the thing that reports back; there’s a celery subtask that runs when the import task completes. [21:02] maas just barfs on us if they're not [21:03] allenap: ok. so is there a way to route the whole thing through an http proxy if i call it through the maas api? [21:04] alternatively, can i just call the celery job manually? [21:04] tych0: If you’ve set the http proxy in MAAS, then call through to it via the API, it should dtrt. [21:04] hmm [21:05] is there an easy way to set the proxy via the cli? [21:06] Should be, /me looks. [21:06] ok [21:06] no worries, i can look [21:06] just iddn't know if you knew off the top of your head [21:06] (allenap) [21:08] tych0: bin/maas foo maas set-config name=http_proxy value=http://... [21:08] cool [21:08] thanks! [21:08] It’s good to refresh my memory :) [21:14] * allenap signs off for the night. [21:28] allenap: hmm [21:29] we might actually want to be notified if the job fails ourselves [21:29] allenap: is there a way we can get some kind of failure notification? [21:42] allenap: so set-config doesn't work [21:42] http://paste.ubuntu.com/7200708/ [22:17] tych0: maas maas maas set-config ... [22:35] ah [22:35] :-( === CyberJacob is now known as CyberJacob|Away