[01:46] bigjools: if i say the words "curtin" and "arm64" or "armhf" in the same sentence does that provoke any thoughts? [01:46] scary ones [01:46] heh [01:47] I don't know if it's ever been tried on those arches [01:47] arm IO mean [01:47] gah can't type today [01:47] the thing that's lurking in my mind is the need (on current platforms) to run flash-kernel [01:47] if you can find an image, give it a go :) [01:47] what sort of images does curtin usually install? cloud images? [01:48] they're big tarballs IIRC [01:48] bigjools: if i say the words "curtin" and "documentation" in the same sentence... [01:48] wheeeeeee [01:50] http://bazaar.launchpad.net/~curtin-dev/curtin/trunk/view/head:/examples/basic.yaml is just installing a clout root.tar.gz [01:50] *cloud [01:52] i didn't think those had kernels in though [01:54] the kernels are downloaded separately [01:59] oh! [02:00] bigjools: do you know where it gets them from? [02:01] mwhudson: http://maas.ubuntu.com/images [02:02] http://maas.ubuntu.com/images/ephemeral-v2/releases/streams/v1/com.ubuntu.maas:v2:download.json [02:02] ah ok [02:04] bigjools: do you know where in curtin (which 'stage') the kernel gets handled? [02:04] I don't, sorry [02:05] it's a bit of a black hole to me, and as you note there's no docs [02:05] i bet smoser doesn't have anything else to do this cycle [02:07] one way appears to be more or less "chroot /target apt-get install linux-image" [02:07] which ought to work fine x-platform [02:09] yeah [02:09] * bigjools brb [02:09] actually i think that's more or less what it always does, not sure it uses the simplestreams data [02:10] oh hahaha [02:10] machine = platform.machine() [02:11] if machine.startswith('armv7'): [02:11] update_initramfs(target) [02:11] else: [02:11] setup_grub(cfg, target) [02:11] yeah so about that... [02:11] not sure why the update_initramfs part is necessary, that's done by installing the kernel pacakge [02:11] unless triggers are disabled or some such insanity [02:15] does the kernel package get installed on a curtin install? [02:15] it's not part of the image that gets unpacked? [02:15] jhobbs: that's the way i read it [02:16] http://bazaar.launchpad.net/~curtin-dev/curtin/trunk/view/head:/curtin/commands/curthooks.py#L139 [02:19] i added that line; iirc the initramfs wasn't getting run until that setup_grub() ran [02:19] which didn't work on armhf, so i just did update_initramfs [02:19] huh [02:20] maybe there is some conditionality then; the kernel package is only installed if there is no kernel in the installed image or something [02:20] but i don't see that [02:27] now [02:27] can i remember why i was interested in this question? [02:36] jhobbs: so the way maas uses curtin is that it netboots a cloud image and uses cloud init data to run curtin? [02:36] how does the cloud init data get to the machine? [02:38] mwhudson: wget from maas more or less [02:38] ah so the cloud-init user-data is "get more config from this url"? [02:40] ah yes [02:50] ah i think you can put something like ds=nocloud-net[s=$url] on the kernel command line [02:50] and then cloud-init will hit $url/user-data and $url/meta-data === CyberJacob|Away is now known as CyberJacob === vladk|offline is now known as vladk [07:11] bigjools: We don't really support multiple IPs per NIC for cluster interfaces, but I figured we could either say TFB, or treat them as multiple interfaces. [07:11] Depending on whether any actual need arises. [07:11] jtv: treat them as aliased separate interfaces [07:11] I did think about it at the time [07:12] I guess we could only manage one anyway. [07:12] Doesn't make much sense to run two dhcpds (for the same IP version) on the same NIC but for different networks. [07:12] (Not counting VLANs where you've got, from our perspective, different NICs anyway) [07:16] no, you'd run a single dhcpd for all interfaces like we do now [07:17] Same thing applies regardless — dhcp daemon or dhcp service, you can't just serve two unrelated dynamic IP ranges on the same NIC without some way to configure which client should go in which range. [07:18] We neither have nor, to my knowledge, need such configuration. :) [07:18] cheap karma: https://code.launchpad.net/~julian-edwards/maas/ui-fixes/+merge/223506 [07:19] jtv: right, the whole thing is somewhat nonsensical [07:20] Cheap karma? Coming! [07:23] bigjools: haha, I see the wrong capitalization came from Django generating the labels based on the name of the fields. :) [07:23] rvba: corrrrrect [07:23] rvba: I am learning too much about Django lately. [07:23] *twitch* [07:24] bigjools: too late, I'm anointing you "APAC Django expert attaché to the MAAS team" ;) [07:24] * bigjools sets up mail forwarder to rvba for anything with "django" into [07:25] s/into/in it/ [07:25] * bigjools having a bad typing day [07:35] Recapitalising words is one of those things frameworks really oughtn't be doing for us. [07:35] bigjools, did you run into one of those "Ip address" fields? [07:35] jtv: I bumped my nose on it === vladk is now known as vladk|offline [07:36] Exercise: auto-capitalise MAISON DU THE for display as a title. [07:37] "Oh, ‘the’ is an article, let's lower-case that." [07:37] (I believe it should become thé) [07:37] (Sorry, Thé) [07:38] jtv: shouldn't it by "MAISON DU THÉ" in the first place? [07:40] In principle, yes — but aren't all-caps names like this often displayed without the accents? [07:40] Very often indeed. But it doesn't make it right. [07:41] But it's easy to think nobody will care, when actually some piece of software may come along and make assumptions about how your capitalisation rules work. [07:42] And don't get me started on how the Scots are going to fill out fields labeled "Mac"... [07:42] "Ach, I entered ‘Gregor’ but it doesnae work!” [07:53] jtv: care to review a tiny branch? https://code.launchpad.net/~rvb/maas/fix-manpage-generation/+merge/223511 [07:53] OK [07:53] Ta === CyberJacob is now known as CyberJacob|Away [08:59] jtv: I also want to get started on doing some actual coding on IPv6. Could we have a call (it will be short) to coordinate our work on this? [09:01] rvba: sure — there's some low-hanging fruit. [09:06] rvba, jtv: please make sure you leave something obvious for me to do [09:06] since I won't be able to pre-imp unless you volunteer to stay up late :) [09:06] bigjools: there'll be stuff left... any questions about the cards on the board? [09:06] bigjools: did you talk with Andres about the additional work he wanted to get done on the DHCP stuff? === vladk|offline is now known as vladk [09:07] jtv: I've not looked yet [09:07] rvba: sort of [09:08] bigjools: they all have detailed descriptions... if you see something unclear about one, it's probably going to be unclear about all of them, so let me know. [09:08] jtv: ok ta [09:09] jtv: I guess the first thing is to try and get ipv6 working on my test network [09:09] There are also a bunch of known jobs that you can do without though. [09:12] jtv: why a feature flag? [09:15] bigjools: because there are places where it doesn't look as if we can do both IPv4 and IPv6 — we'll have to choose. [09:15] jtv: example? [09:15] Which do we tell dhcpd to serve? [09:16] isn't that automatic depending on what format IP addresses are entered in the cluster interface edit form? [09:16] Which cluster addresses do we discover? [09:16] all of them, like now [09:16] We don't do that now. [09:16] We only discover the IPv4 ones. [09:17] that's all of them then, barring the ones we are about to add [09:17] ... [09:18] we already discover all ipv4, we just add discovery of ipv6 interfaces. [09:18] and if you set them to be managed, we tell dhcpd to serve on there [09:19] Yes, we can do that one dynamically. [09:20] We'll need some validation changes to make it work that way of course. [09:23] The problem is that right now we have a bunch of places that will quietly give wrong answers if we mix IPv4 and IPv6. A feature flag buys us the freedom to fix those without breaking normal operation. [09:23] * jtv steps out for about 2 minutes [09:25] jtv: I say just do it right up front [09:26] a feature flag is just procrastinating [09:30] bigjools: I think we can do it without a feature flag, if we can continue to run celery and such as IPv4, assuming the DNS stuff works out. [09:49] MaaS region controller with 2 Cluster | http://askubuntu.com/q/484951 === vladk is now known as vladk|offline === vladk|offline is now known as vladk [10:40] jtv: I think that would be OK from what I read === vladk is now known as vladk|offline === vladk|offline is now known as vladk [11:55] gmb: just spotted one of my pet antipatterns in update_mac_cluster_interfaces. When you find yourself writing a search loop, extract it. Don't do what you need to do inside the loop body and then decide that you're only going to do it once, so you can return. [11:55] "Find item. If found, process item." [11:56] Not: "Loop over items. If item is what I want, process item. Break out of loop as an optimisation." [15:47] anyone here do much with MAAS? I have a fresh install of 14.04 installed as MAAS controller that is being cranky and refuses to load the images. Installed OS, booted up, apt-upgrade, reboot, create MAAS user, login, click download images. Also tried "sudo maas-import-pxe-files" after as a backup and that completes but does not result in the webpage recognizing the images' existence. Also checked [15:47] celery.log and it doesnt spit any errors/warns. [15:49] Solution-X: and the boot images do not show up in the UI? [15:50] It should normally be the one or the other — you get the images, or there's going to be an error in that log... [15:50] (Be aware that the download is huge...) [15:53] correct [15:53] which is whats confusing me [15:55] In the celery.log, do you see the import task at least starting and finishing? Or only starting, or neither? [15:55] give me a moment and ill drop a paste in...there may be something im simply not seeing [15:56] OK. I'll have to leave soon, but would like a look. [15:56] just started and saw the start task and iftop is showing traffic [15:57] Then that sounds as if it's probably still downloading. Have a look in /var/lib/maas/boot-resources/cache; you should see the files growing. [15:58] that is my impression as well [15:59] You should see a snapshot directory in /var/lib/maas/boot-resources, too. [15:59] im trying to more or less tail the process to you without shoving all the logs down your throat [15:59] cache snapshot-20140618-092857 snapshot-20140618-095234 snapshot-20140618-100031 snapshot-20140618-100657 snapshot-20140618-100932 snapshot-20140618-115549 [16:00] Looks like you're running a bunch of downloads at the same time, or maybe started a few that were aborted. [16:00] probably aborted [16:01] Hope so, because if they're still running they may slow you down. [16:01] iftop traffic has subsided, load is 0 https://privatepaste.com/a98fcd3f02 [16:01] bad formatting but that link is celery [16:02] It goes through phases; it may be unpacking an image file. [16:02] Yup, your imports are probably still running. [16:04] If you're running maas 1.5 (the default for 14.04), make sure you only have one instance of the import script running. If the downloads are too much, review /etc/maas/bootresources.yaml. (Finicky syntax, so be careful). [16:04] That file is no longer there in 1.6, but in 1.5 it'll help you restrict what gets downloaded. [16:05] If the script was going through an unpacking phase just now, watch for changes in the latest snapshot directory. [16:05] Gotta run now! [16:05] * jtv runs [16:39] looks like one of the previous tasks was causing the subsequent imports to fail...maybe i click import when it was already attempting an import when i first set it up...cleared the snapshots and ran another import, this time it was successful [16:39] thanks again for the help and advice! [16:53] Clearing the snapshots shouldn't make much of a difference, but they can pile up a bit. They won't use much space though, because it's all links into the cache dir. === matsubara is now known as matsubara-lunch [17:20] Is there a way to turn off "maas maas nodes accept-all" === vladk is now known as vladk|offline === CyberJacob|Away is now known as CyberJacob === deegee__ is now known as drussell === matsubara-lunch is now known as matsubara [18:20] is maas 1.3.1 the latest packaged for ubuntu 12.04.4? [18:21] will i still be able to use juju with this version? [18:26] is there a simple way of installing the latest version of maas on precise? === vladk|offline is now known as vladk === roadmr is now known as roadmr_afk [19:52] Back working on my system that never Commissions. I can now look at the install logs, but I'm not seeing the problem. [20:03] The only problems I'm find are possibly not problems [20:03] Could not open device at /dev/ipmi0, so Unable to start ipmievd during installation. [20:03] and ci-info: no authorized ssh keys fingerprints found for user ubuntu. [20:05] Anyone have troubleshooting suggestions? === lazyPower is now known as lazypower-travel === vladk is now known as vladk|offline [20:29] My cloud-init.log does show an awful lot of "Failed at attempted import of 'modulename'" -- http://pastebin.com/1daAfrTp === roadmr_afk is now known as roadmr === roadmr is now known as roadmr_afk [20:56] I suspect I have this problem which is possibly hardware related - https://bugs.launchpad.net/ubuntu/+source/maas/+bug/1321885 [20:56] Ubuntu bug 1321885 in openipmi (Ubuntu) "IMPI detection and automatic setting fail in ubuntu 14.4 maas" [Undecided,Confirmed] [20:56] I'm going to manually build and curse the time I've wasted [20:58] MilesDenver: can you test running this script on your server [20:58] http://bazaar.launchpad.net/~maas-maintainers/maas/trunk/view/head:/etc/maas/templates/commissioning-user-data/snippets/maas_ipmi_autodetect.py [20:58] if you could run that and post the output to the bug that would be good [20:59] hmmm. ok [20:59] you'll have to isntall freeipmi-utils first [20:59] and run it as root [20:59] that's the script that does ipmi config/detection [20:59] unfortunately it's hard to get good output from it from the logs [21:03] Where does it output? [21:04] it just outputs to console [21:04] when you run it yourself [21:04] hmmm, it only output "maas,WOzMJ8pfeG,172.26.0.186,LAN_2_0" [21:05] well, that's a good sign [21:05] can you use those credentials to access the system? [21:05] oh... with some kind of IPMI tool. [21:05] ah yeah [21:06] from a remote system, you could use ipmitool - ipmitool -H 172.26.0.186 -U maas -P WOzMJ8pfeG power status [21:06] can you also post the output of ipmi-locate ? [21:09] thanks - ipmi-locate http://pastebin.com/Zjtt8s80 [21:10] and the ipmitool responds "Error: Unable to establish LAN session Unable to get Chassis Power Status" which likely means a firewall problem [21:12] MilesDenver: yeah it looks like detection and configuration is working [21:12] MilesDenver: can you ping that host from the maas server? [21:12] I'm supposed to be building an OpenStack Compute Node on this system. Tomorrow I need to build it manually [21:13] ah [21:13] out of time for debug then? [21:13] jhobbs well, I can probably debug a bit longer [21:14] err, can you try this command ipmitool -H 172.26.0.186 -U maas -P WOzMJ8pfeG power status [21:14] oops [21:14] ipmitool -Ilanplus -H 172.26.0.186 -U maas -P WOzMJ8pfeG power status [21:16] jhobbs: sure.... long wait (probably my firewall) then [21:16] "Error: Unable to establish IPMI v2 / RMCP+ session Unable to get Chassis Power Status" [21:16] k [21:16] yeah i agree - sounds like a network issue [21:16] or a wedged BMC [21:17] lifeless: it's network. I've never asked for the ports to be opened. [21:17] MilesDenver: oops;) [21:20] Seems like there is beer downstairs. I'll come back to this tomorrow, but we have an all hands meeting in 10 minutes. [21:21] Actually, I'll probalby end up on the patio with my laptop in about an hour [21:21] sounds rough [21:21] wish you were here === roadmr_afk is now known as roadmr === CyberJacob is now known as CyberJacob|Away