[00:00] if you alt-f8 (I think) you should see more output [00:00] it takes a while [00:00] it downloads some stuff from the repos [00:01] it's been about 30 min. [00:01] ok [00:01] it should be around 5-10 [00:01] alt-f8 didn't give me anything -- but then again, i'm using it over the bmc and the keystroke may or may not be getting across [00:01] ok [00:01] one sec [00:02] i don't suppose you know the default username/password? [00:02] there isn't one [00:02] (let me guess -- password is generated?) [00:02] but there's a backdoor image that you can use to commission with that has a user/password set up [00:02] * bigjools looks for it [00:02] well i can't ssh w/ the key i provided in my preferences [00:02] no, you can't log in at all with commissioning [00:03] it's past commissioning [00:03] it's not - otherwise it would be ready [00:03] it's commissioned -- it's deployed [00:03] ... [00:03] that is, the OS is installed, cloud-init reported as having completed [00:03] ok let's get the terminology straight [00:03] lol okay (c: [00:03] * mwhudson was going to say :) [00:04] you were previously at the "ready" state? [00:04] i was at whatever the name for the state is when you've brought up the machine, maas recognizes it, and then i hit "accept & commission" (or "start & commission") [00:04] that is, it enlisted, powered off, you clicked "accept and commission" and it powered up, did stuff and powered off again [00:05] that ran -- the OS installed, it restarted, there's a login prompt, w/ some output indicating that cloud init completed [00:05] wait [00:05] how did you tell the os installed? [00:05] it probably didn't [00:05] i watched the os install [00:05] on the second boot, did it power down? [00:06] it is not installing anything when it commissions [00:06] nothing usable for you anyway [00:06] honestly, there were a few reboots and i'm not sure when it powered down and when it didn't [00:06] yes, i know [00:06] what state does it say that node has in maas now? [00:07] it starts up and you throw something on there (enough to communicate w/ maas and provide it w/ information so the admin can review and decide to actually proceed) [00:07] what state does it say that node has in maas now? [00:07] and then when you accept and commission, it reboots, and at that point the OS is actually installed [00:08] status shows "allocated to sysadmin" [00:08] But wait if I get it right. If I want MAAS to have another node I just boot it up ones from PXE and MAAS then handles it from then on? [00:08] ok so your story of events is incorrect somewhere [00:08] i'm telling you exactly what i saw [00:08] (as i recall it, granted :D ) [00:08] TheChistoso: if it says "allocated" you clicked something else [00:09] can you "ssh ubuntu@" [00:09] because you've commissioned it, it shut down, and now you clicked "start" [00:09] yes -- i did forget that step [00:10] right, so you have an installed node [00:10] so is it possible that the node is in fact ready for use? [00:10] so ssh to the ubuntu user [00:10] My node can't be booted up from the network (at least it doesn't seem to work) or another way. What could I do about that? Could the node still work with MAAS [00:10] i can't ssh into the machine [00:10] Marlinc: without control of PXE/DHCP you cannot use maas [00:11] PXE and DHCP work [00:11] TheChistoso: how are you trying to ssh in? [00:11] It cant start from poweroff state [00:11] Using WOL [00:11] Thats what I mean [00:12] Marlinc: ok so just power up manually after maas requests a startup [00:12] Okay great [00:12] (in that order, specifically) [00:12] if you try to boot a node that maas doesn't think should be starting, it'll do weird stuff [00:13] from the maas controller, logged in as "sysadmin" (w/ the same account name configured in maas as a super user and w/ the SSH key registered there that corresponds w/ my unix account), i type: ssh ubuntu@ [00:14] i've also tried it on port 2181... [00:14] did you add your public ssh key to the sysadmin account in maas *before* starting up the node? [00:14] fwiw, juju makes this stuff much easier [00:14] of course i can't be sure that's the ip of the machine b/c i can't configure it to use a static ip [00:15] no, you can't because MAAS manages the DNS [00:15] i'd love to use juju w/ this -- but i can't get nodes working in order to use juju w/ it (c: [00:15] you're looking at the problem incorrectly [00:15] you have got nodes working, you just think you don't [00:16] it's a cloud-like resource, not an installer [00:16] once nodes are at "ready" you just set the API key in environments.yaml etc and then you can use juju [00:17] are you using maas to manage DNS? [00:17] Should a MAAS node have PXE boot as primary boot location? [00:17] yes maas is managing dns [00:17] Marlinc: you mean in the BIOS? yes [00:18] Okay [00:18] i understand how it's supposed to work [00:18] TheChistoso: ok so just resolve through there and ignore the IP - the DNS entry will be the same as the node's name in the MAAS UI [00:18] the zone file will be named after the cluster's name [00:20] well nslookup doesn't find it - looking at resolv.conf shows that you're not setting up the local machine to use itself for DNS name resolution [00:22] TheChistoso: no, you have to do that [00:23] maas cannot set up resolv.conf for machines it doesn't boot, it's set via DHCP [00:24] it can set the machine it's running on's network interfaces [00:24] i'm doing these commands from the same machine that maas is running on [00:24] that machine is not managed by maas [00:25] at any rate -- i added 127.0.0.1 to /etc/network/interfaces, restarted networking, and now i can ssh into the machine [00:25] you would think that maas itself would want to use its own DNS services (c: [00:25] I just explained why it's how it is [00:26] i understand what you said -- but i simply disagree that maas shouldn't use its own DNS service to resolve names locally [00:27] I think you don't understand [00:27] but it;s working so goofd [00:27] good [00:27] well please explain -- if i truly don't understand something, then a little education would be good (c: [00:28] I explained above [00:28] "maas cannot set up resolv.conf for machines it doesn't boot, it's set via DHCP" [00:28] yes i understand all of that [00:28] i think i would have been pretty upset if maas had started mutilating the networking of the host i installed it on [00:28] so you understand why you need to set resolv.conf yourself? [00:29] but when you install maas, you could configure the local machine to resolve against itself [00:29] (even if i was using maas-dns, which i'm not) [00:29] at the very least, it would make sense to document this [00:29] maas should not mess with networking on non-managed machines [00:29] I agree with the doc [00:29] we should change it [00:30] since you had trouble working this out [00:30] it's certainly not immediately obvoius that the reason why something can't transition to ready is that nameservers aren't configured properly [00:31] (assuming this is the ultimate culprit) [00:31] you are using the state names incorrectly [00:31] ready is when something is available in the pool [00:31] not something that's deployed [00:32] and nameservers configuration has nothing to do with state transitions [00:32] TheChistoso: there is no state transition in maas when the node is installed and ready for use [00:32] the maas ui shows the node's deployed -- so what's the next state called that i want to get it to? [00:32] I think MAAS isn't a toy but really it feels like it. Especially because I've never worked with PXE and never with machines that boot automatically when needed. Really this is just awesome [00:32] it's just "allocated" [00:32] I just connected another machine tot the network. Set PXE as default boot mode and enabled WOL [00:33] And it works [00:33] it transitions from "ready" to "allocated" when someone requests a node [00:33] mwhudson: the first screen (i have just 1 node) says "1 node deployed" [00:33] Marlinc: great! [00:33] TheChistoso: look at the detail node list [00:33] the entire pie chart is blue [00:33] yes -- that says "allocated" [00:33] so you need to get your verbage consistent [00:33] the pie chart is geared to managing 1000s of nodes at once [00:33] eh, maybe the node list and the pie chart could use the same words... [00:34] yeah, that's inconsistent [00:34] * bigjools files a bug [00:34] Wow WOL is just magic [00:35] Marlinc: WoL is a hack, you can't power down using it [00:35] Well okay thats true [00:35] it suffices for basic testing [00:36] okay, so what would my next steps be? [00:36] i've defined .juju/environments.yaml and i ran juju bootstrap and it tells me that no matching nodes are available [00:36] Mmm the second machine doesn't do much but I'm thinking it is a hardware problem because it is a old computer [00:37] TheChistoso: I presume you only have one node commissioned and it's already in use? [00:38] by "in use" do you mean allocated to a user? then yes (although i'm fuzzy on how a node is allocated to a user -- it is tied to the user that accepts, commissions, and starts the node?) [00:38] TheChistoso: no, it's the user who "starts" it [00:38] (*is it) [00:38] okay -- that makes sense [00:38] only admins can accept for commissioning [00:39] (minor feature request, although perhaps it already exists -- auto-accept, commission, and start, assigning it to a default user) [00:39] that won't be accepted as a feature [00:39] it's not designed to work like that [00:39] you need to think cloud-like [00:40] It looks like the DNS is not working the way its supposed to work: 2013-05-02 02:39:13,994 ERROR Invalid host for SSH forwarding: ssh: Could not resolve hostname fqehy.master: No such file or director [00:40] It is managed at least it should be [00:40] My laptop is using the MAAS server's DNS server [00:41] where's that error from? [00:41] Juju [00:41] since these are meant to serve physical machines, makes sense to me that if you've racked, cabled, and powered it up, then you want to commission it. [00:41] When I try to run juju status [00:42] TheChistoso: the commissioning process is more complicated than that in a DC, often people will want to do burn-ins, update firmware etc. [00:42] Marlinc: can you manually look up that host? [00:42] Nope [00:42] Or well [00:42] It doesn't work [00:43] well my suggestion is for those people who don't want to babysit the process :D [00:43] can you check to see if it's present in the zone file [00:43] Let me take a look [00:43] TheChistoso: we'll add a mass-accept at some point [00:43] that would be better :D [00:43] Where's it stored? [00:43] Marlinc: under /var/ib/maas IIRC [00:44] okay, so do you think that i'm at the point where i should be able to use juju? [00:44] TheChistoso: yep [00:44] you need to release that node [00:44] how do i do that? [00:44] sadly there's no way in the UI of doing it (don't ask...), you need to use maas-cli [00:44] ah... [00:45] Mm there's nothing related to DNS [00:45] maas-cli does API calls [00:45] can you point this out in the documentation? b/c i did not see it. gotta have my eyes checked... [00:45] TheChistoso: I think we have gaps in the docs [00:45] this seems like a major one [00:45] which bit? [00:46] that to actually use a node w/ juju you have to run a command [00:46] you don't [00:46] ? [00:46] well i'm confused (c: [00:47] juju will use available nodes on maas [00:47] but you don't have any available [00:47] because you started one manually already [00:47] you'll have to slow down there b/c i'm thoroughly confused [00:47] basically - just don't use "Start" in the UI [00:48] i was just going to ask that [00:48] it's completely unnecessary [00:48] juju manages everything [00:48] even the SSH key [00:48] is that in the docs as well? [00:48] it's in juju docs I expect [00:48] it's natural, i think, to hit a "start" button when you see it (c: [00:48] but there's a section on juju in the maas docs IIRC [00:48] i don't recall reading anything that says "don't press that button!" [00:49] I think the problem here is that you've not really understood how it works from a higher level. Maybe we need a doc section for that. [00:49] minor feature request: add a label or help text (or something) next to the button in maas when it says 'start" to prevent this from happening [00:50] Mmm.. None of the zone files contain any reference to the node bigjools [00:50] i understand how juju is supposed to work (at a high level), but these are small details. c'mon now -- "don't hit start if you want to use it w/ juju"? that seems pretty confusing to anybody starting out. [00:51] Marlinc: ok that means that something went wrong with parsing the dhcp leases [00:51] TheChistoso: I think it's only confusing if you don't understand resource allocation concepts in maas [00:51] Marlinc: please check the celery.log [00:52] Sure [00:52] TheChistoso: which boils down to a higher-level doc that we need to explain these concepts [00:52] given the bad documentation and the fact that i don't work on the product, i don't think it's hard to see why not [00:53] this support stuff is largely done on an "as best" basis for non-paying customers [00:53] i assumed juju worked w/ already running systems, simply orchestrating services atop already-installed OSes [00:53] Okay the're one error in there but it isn't very clear http://pastebin.ubuntu.com/5624546/ [00:53] I think you need to read the juju docs [00:53] your paying customers don't read docs then, eh? (c: [00:53] TheChistoso: no, they pay for consultants [00:53] have you read your docs recently? [00:54] You're welcome for the help [00:54] Marlinc: aha, that's the problem [00:54] i do thank you for your help -- i'm not trying to be ungrateful :/ [00:54] Marlinc: well, an indication of the problem [00:54] but it does seem likely that a newbie (such as myself) has read the docs more recently than somebody working on the product [00:54] True [00:55] TheChistoso: that's very likely [00:55] the product evolves and things get missed, but the vast majority of people don't have problems [00:55] so that experienmced somebody is trying to warn you that it's confusing. [00:56] how do you know that? my previous attempt had me throwing my hands in the air and moving on to other solutions [00:56] iow, you don't know how many people get frustrated w/ the product and leave w/o saying a word [00:57] I know from speaking to many people they are happy with it. There will always be people with problems though - this is a complicated product solving a complicated problem. [00:57] i know bind, dnsmasq, dhcp, cobbler, pxe, tftp, etc. and i found things confusing [00:57] There is a bug report with the same error but it is pretty old and I don't know if its still relevant [00:57] Marlinc: are there log entries talking about leases? [00:57] perhaps those folks have consultants who know the product and have been using it for a while? (c: [00:57] Well in my syslog [00:58] Marlinc: specifically in the celery.log [00:58] [2013-05-02 02:57:33,775: INFO/Beat] Scheduler: Sending due task provisioningserver.tasks.upload_dhcp_leases [00:58] TheChistoso: like I said, it's best effort, like all OSS projects. [00:58] Marlinc: ok cool, one sec [00:58] [2013-05-02 02:57:33,780: INFO/MainProcess] Got task from broker: provisioningserver.tasks.upload_dhcp_leases[****] [00:58] These things? [00:58] Okay nice [00:58] is maas not backed by canonical? [00:58] it is [01:00] so you have real dollars being thrown at it w/ staff dedicating at least some partial amount of their time to it. that's a great thing! now as an open source user of your product, i'm just trying to warn you that your documentation is lacking and you could be turning away users without knowing it. [01:00] Marlinc: can you see your node's MAC/IP in the leases file? [01:01] It is [01:01] TheChistoso: I'm sorry it wasn't up to scratch for you - people will try to make it better [01:01] okay, so what are my options at this point? can i "un-start" (or "un-use") it so i can use it from juju? [01:01] Marlinc: can you look at the celery-region log for errors [01:01] TheChistoso: look at maas-cli [01:02] it has a release command [01:03] There are bigjools http://pastebin.ubuntu.com/5624558/ [01:03] Marlinc: did you install maas-dns? [01:03] and maas-dhcp [01:03] Yes I did it is running because my laptop is using it right now [01:03] And it is using the dhcp too [01:03] weird [01:04] I did need to edit the /etc/network/interfaces file in order to make the DHCP work [01:04] is maas in /etc/sudoers? [01:04] is "rndc" available? [01:04] rndc is available [01:04] But maas is not in the sudoers file [01:04] mmmm [01:05] which version of maas are you using? [01:05] I've used the Ubuntu 13.04 install [01:05] ok [01:06] maas should be in sudoers anyway, which expains the other "omshell" failure [01:06] not sure why it's not there [01:06] Well I don't know either [01:06] Just add it or? [01:07] packaging should have added it [01:07] one sec [01:08] k [01:10] okay awesome -- pie chart is now green showing "1 node queued" and the node list shows its status as "ready" [01:10] so at this point i can "juju bootstrap"? [01:10] Yes [01:10] awesome! [01:10] environments.yaml is setup? [01:11] Or you could get the same problem as me :P the hostname's of nodes are not resolving [01:12] i ran juju bootstrap and the node is now rebooting [01:14] Marlinc: do you have a /etc/sudoers.d/99-maas-sudoers ? [01:14] Marlinc: can you provide more information wrt your environment? network (including network mask), are these machines on the same L2 network, etc. [01:15] his dns is not resolving because the node is not in the zone file [01:15] Ah that is available [01:15] there are a few causes for that [01:15] I don't think it is my network setup [01:16] Marlinc: have you got zone files in /etc/bind/maas/ [01:16] well bigjools has been working on the issue longer than i've paid attention so probably best for me to follow his line of thinking [01:16] named.conf.maas named.conf.rndc.maas rndc.conf.maas zone.1.168.192.in-addr.arpa zone.2.168.192.in-addr.arpa zone.master [01:17] is your node in zone.master? [01:17] there should be a CNAME if so [01:17] Nope [01:17] hmmm -- well it prompted for the partitioning scheme again [01:17] TheChistoso: I'd ask in #ubuntu-server about that, I've never had that [01:20] It is not bigjools :p [01:20] Marlinc: one secx [01:20] okay :) [01:21] Marlinc: http://pastebin.ubuntu.com/5624592/ [01:21] try that and see what output you get [01:21] Okay [01:22] The output is good [01:22] ip->mac mappings [01:22] you see your node? [01:22] Yep [01:22] ok so something went wrong where it writes out the zone file [01:23] let's try: sudo service apache2 restart [01:24] Mmm okay [01:24] Restarted [01:24] it will reload the wsgi for maas which causes a job to rewrite the zone file [01:24] so check the zone file(s) again [01:25] and check celery-region.log to see if anything went wrong since you restarted [01:26] No new output in that log [01:26] And it is still in none of the bind maas files [01:27] hmmm [01:27] can you try renaming the node [01:27] ands watch the log [01:28] Sure [01:29] It did a 'upload_dhcp_release' [01:29] It did a 'upload_dhcp_leases' [01:29] * [01:29] yeah that runs periodically [01:29] Other then that nothing [01:32] arey you looking in celery-region ? [01:32] Well tailed both of them [01:32] [2013-05-02 02:05:34,807: INFO/MainProcess] Task provisioningserver.tasks.rndc_command[6499426a-eb2c-4b30-bfbc-0314632cef2f] retry: None [01:32] None [01:32] Is the last [01:37] I'm not sure what's going on, there are no errors but the file is not getting written [01:38] I'll reboot the thing [01:38] Maybe something is not working [01:42] Now [01:42] [2013-05-02 03:40:27,551: INFO/MainProcess] Task provisioningserver.tasks.rndc_command[6499426a-eb2c-4b30-bfbc-0314632cef2f] retry: None [01:42] None [01:42] Runs without issues [01:43] But there's still no entry [01:49] bigjools, ^ [01:49] yeah# [01:49] trying to think [01:49] someone else will be around soon that may have more ideas [01:49] Okay [01:49] What is soon? Because it is well time to sleep [01:50] For some hours already though [01:50] Usage: time [-apvV] [-f format] [-o file] [--append] [--verbose] [01:50] [--portability] [--format=format] [--output=file] [--version] [01:50] [--quiet] [--help] command [arg...] [01:50] do mei 2 03:50:07 CEST 2013 [01:50] ! [01:51] there's a guy near your TZ and one about to start, so just ping later [01:51] Okay [01:51] I'll see tomorrow [01:51] Good night [01:52] good night [03:04] bigjools: howdy! just wanted to let you know that it seems there's not gonna be an sru anytime soon [03:04] yay [03:05] bigjools: MIR'ing the dependencies is not as easy as it seems, and will probably require TB intervention, meaning, request the TB to provide a exception for those dependencies [03:05] joy [03:05] yep [04:34] Hmm. i changed the IP of my 13.04 maas box and it needs some reconfigureing but I cant figure out what to run or edit. dpkg-reconfigure maas does not seem do do anything on this release. [04:51] Dr{Who}: it's all done in the UI [04:51] edit the cluster [05:48] okay, so i added a new machine to maas. i didn't hit "start node" yet. the machine is currently powered off. i should be able to just use it automatically from juju, correct? [05:48] e.g. i want to just do "juju deploy mysql" [05:48] it picks the right machine, but i don't see it turning on and installing the OS as expected [06:17] any maas+juju users present? [06:44] bigjools: ping [06:44] bigjools: last time I check,the new maas still haven't been backported to 12.04 yet. Has this changed since? [06:44] few weeks ago [07:04] so, uh, anybody? :D [08:06] Hello [11:51] My Juju only finds one node. When I try to add a unit it says "ProviderError: No matching node is available." there is another node ready [11:51] What could I do about this? [11:56] bug 1064291 perhaps? [11:56] bug 1064291 in juju "Default constraints make no sense on MAAS" [Undecided,New] https://launchpad.net/bugs/1064291 [11:57] But that's only one possibility [11:57] I don't remember how to tell juju to use no constraints at all, but if you find out please add it to the bug. [11:57] It'd be nice to document the workaround. [11:58] Ah so.. I've specified i386 and now it works [11:59] But let me guess [11:59] Now only 32 bit nodes will work [11:59] i m not sure how to remove a global constraint for arch, but for the node name, i use juju "set-constraints maas-name=" [11:59] You can change the default constraints [11:59] then i double check it s set to NULL again with juju get-constraint [12:01] Marlinc: please mark bug 1064291 as "affects me too" [12:01] bug 1064291 in juju "Default constraints make no sense on MAAS" [Undecided,New] https://launchpad.net/bugs/1064291 [12:01] Done [12:01] Thanks! [12:01] So it looks like I'm not the only one [12:03] Is it possible to allow multiple architectures using a constraint? [12:04] I'm not sure. #juju might have more people who can help. [12:43] MAAS should boot a machine when Juju requests it right? [12:45] Correct. Not only boot, but install. [12:45] (thus it takes a while) [12:47] Well the last works [12:47] But it is no longer starting the machine [12:49] Ah now it works [12:55] There's a little issue I think [12:56] One my machine's has needs a binary blob for the NIC. It can't boot connect without so it cannot install fully automatically [13:00] I don't think maas will be able to support that case any time soon. You should be able to hack it though, by customizing the preseed. [13:00] Although it wohn't be able to fetch the preseed. Hmm. [13:01] It can still be done. Perhaps by hacking the installer initrd. Perhaps more trouble than it's worth. I can't think of a non-trivial way to do it off the top of my head. [13:01] Mmm I kept the USB in the machine [13:01] Now it works [13:01]