| === Aaton is now known as Aaton_off | ||
| bigjools | lesson learned: don't expect the development environment to work properly when the Ubuntu package is installed :( | 06:21 |
|---|---|---|
| bigjools | jtv, so does the model for nodegroup look ok? | 06:22 |
| jtv | Looking at it now. | 06:23 |
| bigjools | thanking you | 06:23 |
| jtv | (I don't think you need to define a manager if it's an empty class; you get a close equivalent as the default) | 06:23 |
| bigjools | jtv: the template has it | 06:23 |
| * bigjools shrugs | 06:23 | |
| bigjools | will need one later | 06:24 |
| jtv | On which note, kudos for putting this into its own module from the start. | 06:24 |
| bigjools | :) | 06:25 |
| jtv | How to the ip range and the netmask fit together? | 06:25 |
| bigjools | the netmask is what the DHCP server tells the client | 06:25 |
| bigjools | the range is the IP addresses it can assign | 06:26 |
| bigjools | typically they fit inside the mask | 06:26 |
| jtv | I guess the distinction isn't terribly useful yet, but adding them later would only cause pain. | 06:27 |
| bigjools | they will be needed quite soon actually | 06:27 |
| jtv | OK | 06:27 |
| bigjools | since they are values needed to write DHCP server config | 06:27 |
| jtv | One other thing that jumps out at me is the unique constraints on api_token and api_key. Isn't that something we want to leave completely up to the implementation? | 06:28 |
| bigjools | actually can you check the architecture doc after this and tell me if my progressive enhancement plan looks ok? | 06:28 |
| bigjools | ultimately we want auto-enlistment of new workers | 06:28 |
| jtv | That sounds dangerous. | 06:29 |
| bigjools | how? | 06:30 |
| bigjools | actually don't answer | 06:30 |
| bigjools | I'm not worrying about it yet | 06:30 |
| jtv | On your head be it. :) | 06:31 |
| jtv | By the way, there's a piece of nastiness in Django where a text field can be allowed to be null, but it won't actually give you any sensible behaviour. Instead, the operative question is whether the field will be allowed to be empty. | 06:31 |
| jtv | I think you express that with a constructor parameter “blank.” | 06:32 |
| bigjools | yeah, None equiv to empty or something? | 06:32 |
| jtv | I'm not sure. Once people go down that route, there's no telling how far they'll go. | 06:32 |
| bigjools | the only field that I'm allowing none or blank is "name" | 06:33 |
| jtv | What I meant is that you're passing null=False for api_key, which to django doesn't seem to mean all that much in this case. | 06:35 |
| jtv | I don't know what the default is, but if it's True then this will happily work around and against your wishes like a sailing boat tacking against the wind, and if it's False then the default for “name” shouldn't work. | 06:36 |
| jtv | But that's all just little details. I don't see any real problems off the top of my head. Where's the plan you wanted me to look at as well? | 06:38 |
| bigjools | https://docs.google.com/a/canonical.com/document/d/1PtFKcizW3bmP3QBv771tSGMO4VaCiLkWlr-oDzn-FqU/edit?pli=1# | 06:38 |
| bigjools | see page 3 | 06:39 |
| bigjools | basically I want to have a module that generates DHCPD config | 06:39 |
| bigjools | and we can re-use it in a standalone excecutable for the single worker case | 06:39 |
| bigjools | or on the worker as a Task when required | 06:40 |
| jtv | Reuse ftw | 06:40 |
| bigjools | ok | 06:41 |
| bigjools | now Firefox, why won;t you change tabs when I click on one? | 06:41 |
| jtv | Vell, ze first sing ist zat zey must _vant_ zu change. | 06:42 |
| bigjools | zey clearly don't | 06:42 |
| jtv | Interesting. Vy don't you start by telling me about your mozzer? | 06:42 |
| bigjools | my mozzer bites? | 06:43 |
| jtv | Bites? I hope not! | 06:44 |
| jtv | If your mozzer vere like a dog, zen ssientifikally you must be like a dog also, nein? | 06:44 |
| bigjools | ah FF had decided to put a "do you like this cookie?" dialog *underneath* everything else | 06:45 |
| bigjools | sigh | 06:45 |
| jtv | On a note unrelated to your mother, we're getting rather a lot of the mention-an-imported-identifier-so-it's-not-considered-unused pattern. Maybe we should just define an empty function with a name like never_mind()? | 06:46 |
| bigjools | mozzer bites, from mozzer fliez | 06:46 |
| jtv | ? | 06:46 |
| bigjools | zey bite zu and zuck ze blud | 06:46 |
| bigjools | jtv: ideally I'd like flake to stfu | 06:47 |
| bigjools | there's a directive to make it ignore those iirc | 06:47 |
| jtv | That's the idea: you call a function whose name tells the reader “this is unused but that's fine,” and flake shuts up because you're doing something with the identifier. | 06:48 |
| jtv | def not_actually_used(*args): | 06:48 |
| jtv | """Arguments are deliberately left unused. Pass them to this function to suppress lint warnings about it.""" | 06:48 |
| jtv | Now you can suppress the warning for foo by calling not_actually_used(foo) | 06:49 |
| bigjools | ugly hack though :() | 06:51 |
| jtv | I don't think so — it's just giving an explicit name to existing practice. | 06:52 |
| jtv | Call it suppress_warning_if_unused if you prefer. | 06:52 |
| jtv | You don't need to care how it's implemented when you call it. And the implementation as such will be literally as clean as any function's! | 06:53 |
| bigjools | flake needs to be fixed really | 06:54 |
| bigjools | if it's in the __all__ then don't complain | 06:54 |
| bigjools | jtv: please bless https://code.launchpad.net/~julian-edwards/maas/node_group_model/+merge/108678 | 06:55 |
| bigjools | well, shall I do that hack first | 06:55 |
| * jtv casts his sacerdotal eye over https://code.launchpad.net/~julian-edwards/maas/node_group_model/+merge/108678 | 06:56 | |
| bigjools | it's not changed since you last looked, I just added a commit msg :) | 06:57 |
| jtv | And you added NodeGroupManager to __all__, which seems redundant. | 06:58 |
| jtv | Or maybe you did that before. | 06:58 |
| bigjools | it was there before | 06:59 |
| jtv | It probably should not have been. We have one canonical way of referring to the manager. | 07:00 |
| jtv | Hi rvba | 07:00 |
| bigjools | jtv: ok | 07:00 |
| * jtv growls at django docs | 07:00 | |
| bigjools | hello roaksoax | 07:00 |
| bigjools | errr hello rvba | 07:01 |
| jtv | specifically this part: «When using the Oracle database backend, the null=True option will be coerced for string-based fields that have the empty string as a possible value, and the value NULL will be stored to denote the empty string.» | 07:01 |
| jtv | So the “null” parameter defaults to False, except on one database which doesn't really do nulls anyway. | 07:01 |
| jtv | Heyyy, there's rvba again :) | 07:01 |
| rvba | Hello bigjools. Hi jtv. | 07:01 |
| jtv | bigjools: you have my blessing. Go forth and sin some more. | 07:07 |
| bigjools | jtv: thank you my child | 07:07 |
| jtv | Or at least, that's what it always sounded like to me. I have may have gotten some of the details wrong. | 07:07 |
| bigjools | or is that, thank you father | 07:07 |
| * jtv is reminded of Blackadder and his father in the archbishop episode | 07:08 | |
| jtv | rvba: I already made the same suggestion about leaving out the default manager class, but I was hoping to use it against Julian next time he thinks my branches are too big. :) | 07:09 |
| rvba | jtv: haha | 07:09 |
| * jtv searches for coffee, which he hasn't had all day | 07:09 | |
| jtv | Hitting the office before the Ozzies wears one out. | 07:10 |
| jtv | bigjools: do we have data that supports the single-DNS design? I thought we didn't. | 07:34 |
| bigjools | jtv: it's largely irrelevant actually | 07:35 |
| bigjools | because it has a scaling solution of its own | 07:35 |
| jtv | Ah | 07:35 |
| jtv | rvba: would it be bad in any way to import django settings in a urls.py? | 07:54 |
| jtv | Argh. Did I break the build? I get UserProfile test failures now! | 07:56 |
| rvba | jtv: In an application's url.py it's ok. It's already done in src/maasserver/urls.py. | 07:59 |
| jtv | OK | 07:59 |
| jtv | Hmm Jenkins seems to blame bigjools for the breakage. | 08:00 |
| rvba | Yep. | 08:00 |
| * jtv stumbles towards coffee to shake off his déjà vu. | 08:01 | |
| jtv | Hmm… I wonder… | 08:03 |
| * jtv tries something | 08:03 | |
| rvba | Hum, all tests pass locally… | 08:03 |
| rvba | Well, I only tried the faulty tests. | 08:04 |
| rvba | But they pass. | 08:04 |
| jtv | Wrong TestCase class, I suspect. | 08:04 |
| jtv | (I'm already running tests to confirm) | 08:04 |
| * rvba runs the whole test suite… | 08:06 | |
| jtv | Don't bother. I've already got a branch that fixes this & a bunch of other small things. | 08:07 |
| rvba | ok, cool, thanks jtv. | 08:07 |
| bigjools | I swear I ran everything | 08:14 |
| bigjools | WTF | 08:14 |
| jtv | Would anyone care to review? Urgent urgent urgent test breakage urgent: https://code.launchpad.net/~jtv/maas/test-before-landing-bitch/+merge/108697 | 08:15 |
| bigjools | looking | 08:16 |
| rvba | jtv: I'll take it. | 08:16 |
| jtv | Thanks. | 08:16 |
| * bigjools wins | 08:16 | |
| jtv | Work it out amongst yourselves. :) | 08:16 |
| bigjools | jtv: also, blow me! | 08:16 |
| jtv | >-O | 08:16 |
| jtv | !? The diff doesn't show all my changes. | 08:17 |
| ubot5 | jtv: I am only a bot, please don't think I'm intelligent :) | 08:17 |
| jtv | Did I commit wrongly? | 08:17 |
| jtv | ubot5: stay out of this | 08:17 |
| ubot5 | jtv: I am only a bot, please don't think I'm intelligent :) | 08:17 |
| jtv | Ah no, this is the whole thing. | 08:18 |
| jtv | Well here we are then. | 10:31 |
| jtv | And I think I brought that same question up in pre-review. :) | 10:31 |
| rvba | bigjools: skip_check is exposed to the user because it's used in the API. | 10:31 |
| bigjools | ah | 10:32 |
| rvba | bigjools: that's the whole point of 'skip_check', it has to be usable by the user to bypass all the checks if he wants to. | 10:32 |
| jtv | Maybe it should even be off by default then... so that nobody assumes that the check always happens. | 10:33 |
| bigjools | rvba: ok it makes more sense now | 10:33 |
| rvba | bigjools: Cool. | 10:33 |
| rvba | jtv: well, I don't think that would be wise, the check is also a guid for the user to provide the right parameters. | 10:33 |
| jtv | I could imagine the validity of a particular power_parameters object changing without that object itself changing. | 10:33 |
| rvba | guide* | 10:33 |
| rvba | jtv: not sure I follow… | 10:34 |
| jtv | Damn. I had something in mind back then and didn't write it down. | 10:34 |
| jtv | There may be scenarios (not in the code now perhaps, but at some point) where you set a node's power parameters, and they are valid; and then you make some other change, and now that node's power parameters are no longer valid. | 10:35 |
| rvba | That can already happen if you change the default parameter and power_type is not set on the node. | 10:36 |
| rvba | the default power_type that is. | 10:36 |
| jtv | Yes, but would you go through the check in that case? | 10:36 |
| jtv | Ah, default power type. | 10:36 |
| rvba | Yes, the check would be done. | 10:37 |
| jtv | What I'm driving at is that, in cases like that, the check provides no guarantee. Which is fine, but it may not always be obvious. | 10:37 |
| jtv | Also, I've seen cases in the past where the logic around such a parameter shifted and grew and its effective meaning changed subtly, and it led to no-go zones in the code. | 10:38 |
| jtv | By which I mean pieces of code you can't maintain effectively, not dead code. | 10:38 |
| jtv | Not saying that any of us would make that mistake, but I'm reasoning from the AK-47 / SA-80 perspective here. | 10:39 |
| rvba | Well, in this instance I think it would be quite easy to maintain that code if the requirements change for this. Mainly because all the logic is fairly well encapsulated in a field. It's really decoupled from the usage of the field. | 10:40 |
| rvba | What I mean is that we could easily create another field with another logic and replace the old field by the new one. | 10:40 |
| rvba | As long as the storage is compatible (JSON blob). | 10:40 |
| jtv | Actually, where does the skip_check parameter go? | 10:42 |
| rvba | It's used by the field to decide whether or not validation should be performed. | 10:43 |
| rvba | It's not stored. | 10:43 |
| jtv | Okay, but where is it used? I see it tested, and I see it documented, but where does its value get read? | 10:44 |
| jtv | I don't see it in the current trunk, nor do I see it being read anywhere in the diff. Prerequisite branch? | 10:45 |
| rvba | In a pre-req branch (where the custom field/widget are defined): https://code.launchpad.net/~rvb/maas/multifield/+merge/108188 | 10:45 |
| jtv | Ah. And its mentioned quite a lot there I see! | 10:47 |
| jtv | Then messing with it would be too much work for this branch anyway, so I'll stop worrying about it. | 10:47 |
| jtv | I think I'll worry about people setting hostnames like “localhost” instead. | 10:48 |
| rvba | bigjools: did you change this branch's status manually? | 10:57 |
| rvba | The 'multifield' branch. | 10:57 |
| bigjools | rvba: I put an email directive in to set my vote to approved | 10:58 |
| bigjools | ah bollocks | 10:58 |
| rvba | bigjools: not sure what happened, it's now merge… even before I had time to commit my changes | 10:59 |
| rvba | merged* | 10:59 |
| bigjools | I set the MP status rather than my vote | 10:59 |
| bigjools | sorry | 10:59 |
| bigjools | I confused "status" and "merge" | 10:59 |
| rvba | bigjools: ok, no worries, I'll fix that in the follow-up branch. | 10:59 |
| bigjools | ok, sorry for the trouble | 11:00 |
| rvba | Well, actually I think the problem will fix itself since I pumped the new revisions into the followup branch already. | 11:01 |
| rvba | hum, no, it doesn't seem to be working… | 11:03 |
| jtv | Good feeling: putting my branch up for review at 18:03. I think I'll be off. | 11:04 |
| rvba | jtv: I'll review it later this afternoon. | 11:05 |
| jtv | Thanks! | 11:05 |
| rvba | nn jtv. | 11:05 |
| jtv | nn! | 11:08 |
| roaksoax | smoser: ok, so I've hit that issue, and I've seen people hitting that issue | 13:08 |
| smoser | roaksoax, so.. the easiest way to recover there is to hav the installer set a password for the ubuntu user | 13:08 |
| smoser | so the person can get into the instance to look around | 13:09 |
| roaksoax | smoser: I can't recall how I managed to "fix" it but I'm guessin it has something to do with DNS being different from what is given, to what it is stored in maas | 13:09 |
| roaksoax | smoser: right, but wasn't that the idea of having the ability to add anSSH key to MAAS? | 13:09 |
| cheez0r | I'm having this issue right now, where I've bootstrapped juju, but get "Invalid SSH key" from juju status, and I have no way (no l/p, no ssh key) to access the system it's running on to troubleshoot/fix the issue | 13:11 |
| smoser | roaksoax, well, yes. | 13:11 |
| roaksoax | smoser: ^^ | 13:12 |
| smoser | but that ssh key is pulled in by cloud-init | 13:12 |
| smoser | and that is failing | 13:12 |
| smoser | which is why you're not getting in | 13:12 |
| roaksoax | smoser: isn't the cloud-init failing due to the timezone being different from maas client/server? | 13:12 |
| smoser | so setting a password is an easy way to fix "and I have no way (no l/p, no ssh key) to access the system its running on to trubleshoot/fix the issue" | 13:12 |
| smoser | roaksoax, well,t hat will fail first in commissioning state | 13:12 |
| smoser | but yes, it is a potential point of failure, and admittedly it could also have such a bad clock that time was lost between commissioning and deployment | 13:13 |
| smoser | (but unlikely) | 13:13 |
| smoser | cheez0r, are you willing to debug a bit? | 13:13 |
| cheez0r | smoser: gladly. | 13:13 |
| cheez0r | I've been stuck on this for a while. | 13:13 |
| roaksoax | pmatulis: in your case, are you guys sstill using the wrt router as DNS/DhCP server? | 13:14 |
| pmatulis | roaksoax: no, all is on the maas server | 13:14 |
| smoser | cheez0r, in /var/lib/cobbler/kickstarts you will have a file named 'maas.preseed' (or somethign to that effect) | 13:15 |
| smoser | (sorry, just going from memory) | 13:15 |
| smoser | then... | 13:15 |
| smoser | you want to set a password in there | 13:15 |
| smoser | d-i passwd/user-password-crypted password $6$.1eHH0iY$ArGzKX2YeQ3G6U.mlOO3A.NaL22Ewgz8Fi4qqz.Ns7EMKjEJRIW2Pm/TikDptZpuu7I92frytmk5YeL.9fRY4. | 13:15 |
| smoser | tha tis 'ubuntu' | 13:15 |
| roaksoax | pmatulis: i see. smoser i think I know what the issue is | 13:15 |
| smoser | clearly you dont want this for production, but after a re-installation, you should be able to log in as 'ubuntu:ubuntu' | 13:15 |
| roaksoax | smoser: both pmatulis and cheez0r are using cobbler's dnsmasq and I think maas is not correctly updating what it has to update | 13:16 |
| roaksoax | smoser: err maas-dhcp | 13:16 |
| cheez0r | s'ok with you if I grab my own crypted password? | 13:17 |
| roaksoax | cheez0r: yeah go for it | 13:17 |
| cheez0r | ok done | 13:17 |
| cheez0r | then re-commission the nodes? | 13:18 |
| pmatulis | roaksoax: i also noticed the server's resolv.conf was populated with 127.0.0.1 for some reason. dns was broken and i had to hand-edit the file to get anywhere | 13:18 |
| smoser | cheez0r, re provision | 13:18 |
| smoser | cheez0r, or, were you failing in commission stage? | 13:19 |
| roaksoax | pmatulis: that's foundationschanges to how resolvconf is being handle now | 13:19 |
| roaksoax | pmatulis: so aaaall dns clients will have 127.0.0.1 | 13:19 |
| smoser | pmatulis, resolv.conf should probably have 127.0.0.1, as you'll be running dnsmasq | 13:19 |
| cheez0r | I'd succeeded with commissioning, the nodes were in "Ready", but when I run juju bootstrap then juju status, I get the Invalid SSH Key error. | 13:19 |
| cheez0r | so I guess just juju destroy-environment and juju bootstrap and see what happens? | 13:19 |
| smoser | cheez0r, yeah | 13:20 |
| smoser | and at the point when it re-createds | 13:20 |
| smoser | then you can try to ssh in as 'ubuntu:ubuntu' | 13:20 |
| smoser | and we can poke around a bit. | 13:20 |
| cheez0r | Also, my nodes are all configured for WOL but for some reason start node in the MAAS GUI isn't doing anything, but we can skip that for now | 13:20 |
| smoser | yeah. | 13:20 |
| cheez0r | I figured out that poweron sidesteps that | 13:20 |
| smoser | WOL is often buggy anyway | 13:20 |
| smoser | but that is a separate issue. | 13:20 |
| smoser | use sneaker power now. | 13:20 |
| cheez0r | No doubt. | 13:20 |
| smoser | :) | 13:20 |
| cheez0r | ILO for the win :p | 13:21 |
| pmatulis | roaksoax, smoser: i'm aware of the resolvconf stuff but somehow i need to configure dnsmasq to talk to my actual nameserver | 13:21 |
| smoser | pmatulis, well, that should happen. | 13:21 |
| smoser | it does potentially point to an issue with dhcp not giving it a dns server | 13:21 |
| pmatulis | smoser: well, my sever has a static address | 13:22 |
| smoser | pmatulis, i don tknow how that would be supported in maas. | 13:23 |
| smoser | but maybe i'm missing something. | 13:23 |
| smoser | oh. | 13:23 |
| smoser | i see. | 13:23 |
| smoser | a non-maas system. | 13:23 |
| pmatulis | smoser: i'm talking about my maas server | 13:24 |
| roaksoax | pmatulis: you should just be able to replace /etc/resolv.conf with a file rather than being a symlink | 13:24 |
| roaksoax | pmatulis: http://www.stgraber.org/2012/02/24/dns-in-ubuntu-12-04/ | 13:24 |
| roaksoax | that would help then ^^ | 13:25 |
| pmatulis | roaksoax: that's true, but is that the proper way? | 13:25 |
| roaksoax | pmatulis: if it is a static addres, then you can specify the dns servers on /etc/network/interfaces | 13:25 |
| pmatulis | roaksoax: i did that already | 13:26 |
| smoser | pmatulis, dns-nameservers 12.34.56.78 12.34.56.79 | 13:26 |
| smoser | http://askubuntu.com/questions/32875/network-manager-discards-changes-to-etc-resolv-conf-on-boot | 13:26 |
| smoser | that is the correct way to configure /etc/resolv.conf with static networking in the presense of resolvconf. | 13:26 |
| roaksoax | yeah | 13:26 |
| smoser | sadly 'dns-nameservers' is not even mentioned in 'man intefaces' :-( | 13:27 |
| * smoser files a bug | 13:27 | |
| roaksoax | pmatulis: stgraber's blog has more info about it too | 13:27 |
| roaksoax | smoser: it is mentioned on man resolvconf though | 13:27 |
| cheez0r | I think you have to man resolvconf to find it | 13:27 |
| pmatulis | roaksoax: well, dns-nameservers is already done | 13:27 |
| roaksoax | yep | 13:27 |
| cheez0r | hey smoser/roaksoax, that worked. I'm in the juju system now. | 13:28 |
| roaksoax | cheez0r: are there any juju logs ? (/var/log/juju/*) | 13:28 |
| cheez0r | didn't even have to re-provision, just reboot the machine | 13:28 |
| cheez0r | yes | 13:28 |
| roaksoax | cheez0r: so you can now juju status? | 13:29 |
| cheez0r | client environment stuff, zookeeper_init, juju.agents.machine Machine agent started | 13:29 |
| roaksoax | cheez0r: wait wait, juju status now works after a reboot? | 13:30 |
| cheez0r | well, I had to fix some ssh issues, but yes | 13:30 |
| cheez0r | the ssh fingerprint changed for some reason | 13:30 |
| cheez0r | now it shows machines: 0: agent-state: running | 13:30 |
| smoser | cheez0r, well, the ssh fingerprint *should* change | 13:31 |
| smoser | you re-installed the system. | 13:31 |
| smoser | oh. | 13:31 |
| smoser | i just saw above. | 13:31 |
| roaksoax | smoser: he didn't | 13:31 |
| smoser | but that is still kind of expected. | 13:31 |
| roaksoax | yeah | 13:31 |
| smoser | cloud-init re-generated your ssh keys i suspect | 13:31 |
| cheez0r | yeah I noticed that in the boot scroll | 13:31 |
| smoser | as it found the metadata service this time. | 13:31 |
| cheez0r | if this cold boots again, this happens again? | 13:32 |
| cheez0r | kind of makes it sketchy for ssh key mgmt | 13:32 |
| smoser | cheez0r, no. only once per "instance" (install) | 13:32 |
| cheez0r | ah, ok | 13:32 |
| roaksoax | cheez0r: you only have 1 network interface in the maas server right? | 13:32 |
| cheez0r | no, two, but I've got routes set up properly and dpkg-reconfigure maas done, as well as maas-dhcp configured for the correct parameters | 13:33 |
| roaksoax | uhmm | 13:34 |
| smoser | cheez0r, do you hchok... so cloud-init on reboot found the maas server. | 13:37 |
| smoser | so we have ot fiturer out why it did not on the first boot | 13:37 |
| smoser | hoepfully that information is in /var/loog/maas (or maybe it goes to /var/log/apache/...) | 13:37 |
| smoser | but we should look at logs there around the timestamps | 13:37 |
| smoser | (or just try to reproduce) | 13:37 |
| smoser | i have no idea what 'do you hchok' meant. | 13:38 |
| cheez0r | me either, so I'll say no, I don't. | 13:38 |
| smoser | :) | 13:38 |
| cheez0r | ok so I was able to successfully deploy mysql and wordpress charms, which is progress. | 13:38 |
| cheez0r | smoser: /var/log/maas on the maas node or the juju system? | 13:39 |
| smoser | maas system | 13:40 |
| cheez0r | lots of data in /var/log/maas/maas.log on the maas node. | 13:40 |
| smoser | want to pastebin it? | 13:41 |
| roaksoax | we would also need the juju logs for the bootstrap | 13:41 |
| smoser | also /var/log/apache/error.log | 13:41 |
| roaksoax | s/for/of | 13:41 |
| cheez0r | hrm, to the end, lots of "No matching node is available" | 13:41 |
| cheez0r | towards the beginning, 'no registered public keys' | 13:42 |
| roaksoax | cheez0r: pastebins would be very helpful :) | 13:43 |
| cheez0r | http://pastebin.ubuntu.com/1025043 | 13:44 |
| smoser | cheez0r, did you look at apache error.log ? | 13:44 |
| smoser | that has been where there were hints before | 13:45 |
| cheez0r | no errors other than favicon.ico | 13:45 |
| smoser | :-( | 13:45 |
| roaksoax | cheez0r: can you please pastebin the juju logs from the bootstrap node? | 13:46 |
| roaksoax | rvba: ping | 13:46 |
| roaksoax | rvba: rvba NodesNotAvailable: No matching node is available. --> how do we check for matching nodes, by MAC address? | 13:46 |
| cheez0r | apache access.log: http://pastebin.ubuntu.com/1025046 | 13:46 |
| cheez0r | juju logs from bootstrapped machine: http://pastebin.ubuntu.com/1025049 | 13:48 |
| roaksoax | cheez0r: | 13:49 |
| roaksoax | cheez0r: there | 13:49 |
| roaksoax | 's ponly 1 log for juju? | 13:49 |
| roaksoax | ah nevermind | 13:50 |
| cheez0r | sorry, should have delimited that better | 13:52 |
| roaksoax | so while there doesn't seem to be an issue with juju for the bootstrap it does seem to be an issue when deploying wordpress and mysql | 13:53 |
| cheez0r | well, only one of the two nodes I have is successfully resolving via DNS. | 13:53 |
| cheez0r | I'm using dnsmasq on the MAAS node to do resolution but for some reason cobbler isn't successfully populating the DNS entries. | 13:54 |
| roaksoax | yeah i need to setup a test environment | 13:55 |
| roaksoax | cause I have no idea what might be wrong | 13:55 |
| cheez0r | hrm | 13:55 |
| cheez0r | second node won't let me log in. | 13:55 |
| cheez0r | let me cold boot it and see if it fixes the same way the other did. | 13:55 |
| cheez0r | I've got to break away for a bit here in a few minutes but let me know how else I can help | 13:58 |
| cheez0r | My ultimate goal is to get openstack running on this system. | 13:58 |
| roaksoax | ok cool | 13:59 |
| roaksoax | cheez0r: i'll try to reproduce locally and see what I find | 14:00 |
| cheez0r | thanks brother | 14:00 |
| cheez0r | the only strangeness about my config is that my MAAS Node has a public IP and a private IP, and that it's got external, upstream DNS servers to resolve against. | 14:00 |
| cheez0r | all of the DNS for within the MAAS is supposed to come from the MAAS Node. | 14:00 |
| roaksoax | exactly | 14:05 |
| roaksoax | cheez0r: so DEFAULT_MAAS_URL = "http://PRIVATE-IP/" so that should be also set in /etc/maas/maas_local_settings.py when you dpkg-reconfigure maas | 14:06 |
| pmatulis | roaksoax: fyi, i redid another node and 'juju status' now gives 'ERROR Invalid host for SSH forwarding: ssh: Could not resolve hostname node5: Name or service not known'. where to troubleshoot name resolution in this context? | 15:52 |
| roaksoax | pmatulis: well the maas server should be the one who knows how node5 is | 15:54 |
| roaksoax | s/how/who | 15:55 |
| roaksoax | pmatulis: are you running juju from the maas server? | 15:55 |
| pmatulis | roaksoax: doesn't know about it. note that i changed the node name within the maas UI | 15:55 |
| pmatulis | roaksoax: yes | 15:56 |
| roaksoax | pmatulis: so 1 would be to check whether the hostname has been updated in cobbler | 15:56 |
| roaksoax | pmatulis: 2, would be to make sure the maas server can resolve the hostnames for the nodes | 15:57 |
| roaksoax | locally | 15:57 |
| roaksoax | 3. that should make juju be able to resolve the hostnames | 15:57 |
| pmatulis | roaksoax: sure, the question is how | 15:59 |
| roaksoax | pmatulis: /etc/resolv.conf add maas server IP as dns server | 15:59 |
| pmatulis | roaksoax: it's already pointing at localhost | 16:01 |
| roaksoax | pmatulis: 127.0.0.1? | 16:01 |
| roaksoax | or 127.0.1.1? | 16:01 |
| pmatulis | roaksoax: the latter | 16:03 |
| roaksoax | pmatulis: what if you make it to be the IP of the maas server | 16:04 |
| pmatulis | roaksoax: sorry, the former | 16:04 |
| pmatulis | 127.0.0.1 | 16:04 |
| roaksoax | pmatulis: make it the MAAS IP address | 16:04 |
| pmatulis | roaksoax: so do not use the resolvconf stuff, use a real file (not symlink)? | 16:04 |
| roaksoax | pmatulis: eitherway, thething is that juju should e able to resolve dns provided *by* the maas server | 16:05 |
| roaksoax | so resolconf provides its own dns with dnsmasq, while maas also provides a dns with dnsmasq | 16:05 |
| roaksoax | so you can setup resolvconf to change the dns server | 16:06 |
| roaksoax | so you should be able to use the same IP addres (for dns) as the MAAS server | 16:06 |
| cheez0r | roaksoax: it is set properly in /etc/maas/maas_local_settings.py | 16:15 |
| roaksoax | smoser: pmatulis cheez0r eok so I think I know what's wrong | 16:27 |
| cheez0r | the suspense is killing me here | 16:30 |
| cheez0r | :p | 16:30 |
| roaksoax | basically, it seems that we might have to specify the IP address for the system because onbviosly, in order to be a dns entry we need to know the IP address for that DNS entry | 16:31 |
| roaksoax | so since dnsmasq has a dns entry but no IP address for a cobbler system, then it doesn't create the entry | 16:31 |
| cheez0r | isn't dnsmasq serving the dhcp IP? | 16:31 |
| cheez0r | or does maas-dhcp install dnsmasq and some dhcpd? | 16:32 |
| roaksoax | cheez0r: right, so here's the deal | 16:32 |
| roaksoax | If dnsmasq has a hostname for Node0, but it doesn't know what IP address to assign yet (as there's being no requests yet), but it knows about Node0 mac address | 16:33 |
| roaksoax | should it resolve Node0 once dnsmasq assigns that IP address? | 16:33 |
| roaksoax | that's my question | 16:34 |
| cheez0r | it should, but what mechanism would update the dnsmasq entry with the IP Address? | 16:34 |
| cheez0r | Perhaps we should be statically assigning dhcp entries so that the DNS can be mapped? | 16:34 |
| roaksoax | yeah that's the common fix | 16:34 |
| cheez0r | Add a static assignment for the MAC into the dhcpd configuration, then populate that assignment into the dns | 16:34 |
| roaksoax | cause, either way, whne you deploy a DNS server you *have* toassign an IP address for that dns name, right? | 16:35 |
| cheez0r | for what DNS name? That of the node? | 16:35 |
| roaksoax | in general terms | 16:35 |
| roaksoax | when you have a DNS solution, in order for you to assign nodeX a hostname being handled by DNS, you need to know the hostname's IP address, correct? | 16:36 |
| cheez0r | right, so a node is commissioned. The MAC is entered into MAAS. The MAC is statically assigned a DHCP IP and the DHCP server is configured as such. The IP and nodename are then pushed into the DNS server. | 16:36 |
| cheez0r | correct | 16:36 |
| cheez0r | that's the only workaround I see unless there's some mechanism called after the node boots and pulls a DHCP IP to then update the dns entry. | 16:36 |
| roaksoax | so yeah the flow in maas should be, once it PXE boots, update MAAS (cobbler really) with the obtained IP address to match dns/ip | 16:37 |
| roaksoax | smoser: ^^ | 16:37 |
| cheez0r | And/or a package designed to watch the DHCP allocation and match up MAC vs. MAC from cobbler and then map the nodename to DNS entry | 16:37 |
| smoser | sort of reading.. wasn't following. | 16:38 |
| smoser | essentially the nodes did not have dns (or reverse dns) entries? | 16:38 |
| === Aaton_off is now known as Aaton | ||
| roaksoax | yeah | 16:39 |
| roaksoax | because cobbler doesn't know what IP address was assigned for that node, so consequently, dnsmasq doesn't know what hostname to match with what IP address | 16:39 |
| roaksoax | smoser: I have a couple ideas of how to fix this. 1. is using something like # Always set the name of the host with hardware address | 16:43 |
| roaksoax | # 11:22:33:44:55:66 to be "fred" | 16:43 |
| roaksoax | #dhcp-host=11:22:33:44:55:66,fred | 16:43 |
| pmatulis | isn't this what ddns is for? | 16:49 |
| === matsubara is now known as matsubara-dinner | ||
| === Aaton is now known as Aaton_off | ||
| === Aaton_off is now known as Aaton | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!