[01:04] actually looks like a CA issue :-/ [02:22] any one can help -> Problem with juju bootstrap [02:28] the command.log is under https://github.com/Ubuntu-Solutions-Engineering/openstack-installer/issues/1042 , and the first lds server is in deployed state under maas... [06:43] is there a way to logon to a failed commission machine in maas [06:44] i need to debug [06:44] i tried ubuntu/ubuntu === frankban|afk is now known as frankban [08:08] Good morning Juju world! [09:10] hi here :) [09:10] lazyPwr: just discovered a bug after 1.5.2 -> 1.5.3 upgrade: [09:10] root@mth-k8smaster-01:~# kubectl exec -it nginx-ingress-controller-0lxzg /bin/bash [09:10] Error from server: error dialing backend: x509: certificate signed by unknown authority [09:10] it just affect the "kubectl exec" subcommand [09:10] all other kubectl subbcommands works with no certificate error [09:10] any clues? [09:11] Hi Zic, thank you for reporting this [09:11] another one: "kubectl logs" is also affected [09:12] as I regularly do other subcommand, I think only exec & logs are affected [09:13] would you be able to open an issue so we can track this down? [09:13] Zic: https://github.com/kubernetes/kubernetes/issues [09:13] yep [09:14] Zic many thanks [09:16] for information, the kubectl logs error is a bit more explicit: Error from server: Get https://mth-k8s-02:10250/containerLogs/development/phpbackend-w0bnv/phpbackend: x509: certificate signed by unknown authority [09:16] so I just curl'ed this [09:17] (openssl s_client -connect mth-k8s-02:10250 actually) [09:18] http://paste.ubuntu.com/24187612/ [09:20] And this is after upgrading from 1.5.2 to 1.5.3, right? I will try to repro today [09:20] and it fact, certs was updated if I observe their date: [09:20] http://paste.ubuntu.com/24187616/ [09:20] kjackal: yep, all charms of the CDK bundle to latest version [09:20] including EasyRSA [09:20] Do you have the issue URL? [09:20] I'm writing it, just a moment :) [09:21] Thanks [09:21] was collecting some more logs to add to it ^^ [09:24] hmm, s/just a moment/I will ping you with the issue's URL/ (because morning do disaster: "GitHub: There have been several failed attempts to sign in from this account or IP address.") [09:24] it seems that I don't know to type my GitHub's password without a coffee [09:25] "Please wait a while and try again later." [09:25] :) [09:41] just got access to my GitHub's account :D I'm preparing the issue :) === frankban is now known as frankban|afk [10:01] kjackal: https://github.com/kubernetes/kubernetes/issues/43209 [10:02] Awesome, thank you Zic [10:03] don't know if I need to tag something because it's about CDK [10:03] (I precised it in the title and "Install tools") [10:05] ok, Zic === frankban|afk is now known as frankban [10:37] morning [10:38] if you add a maas to juju, does juju use your credentials to do things for all users? or does every user needs their own maad credentials? [11:09] Hello there. I'm using maas togher with juju (or the other way arround ;) ). Now i wonder if i want to use the latest LXD version 2.11 instead of 2.8, is that possible with juju? [11:09] cnf: Depends, if you share your juju credentials, then yes. [11:09] cnf: looking here https://jujucharms.com/docs/2.1/clouds-maas each user should/could have its own maas certs that need to be passed to juju [11:10] kjackal: could, or has to? [11:10] kjackal: so a user needs some maas credentials to use the juju command to provision against maas? [11:10] cnf: Yes [11:11] without credentials juju can't connect and control maas [11:11] BlackDex: after bootstrapping [11:14] kjackal: so that means anyone using juju, at any time, needs credentials for the controller, AND the cloud juju is running on? [11:15] cnf: correct [11:15] hmm... [11:16] that's a lot of account-making [11:18] cnf: the benefit is that changing cloud (eg moving from maas to aws and back) is transparent [11:18] how so? [11:20] hmm [11:20] kjackal: so how does the juju webui work? [11:20] that caches credentials, doesn't it? [11:20] I have credentials for aws, openstack, local deployment with lxd and google's GCE. I imported the credentials once and I can switch between clouds with juju switch, everyhing else stays the same (juju deploy kubernetes everywhere) [11:21] kjackal: but those are juju credentials [11:21] kjackal: so if i make a juju controller on AWS, and i make a user on that controller for you [11:21] do you then need AWS credentials to use that controller? [11:26] cnf: have a look here: https://jujucharms.com/docs/2.1/tut-users [11:29] cnf: here is a better explanation on the users in juju https://jujucharms.com/docs/2.1/users [12:09] yeah, been reading that [12:09] it's not entirely clear to me, still [12:12] kjackal: https://jujucharms.com/docs/2.1/tut-users seems to imply juju caches the cloud credentials used during bootstrap, and uses that for all juju users [12:45] Can I introduce new relation on existing charm? [12:46] Meaning that I'm running some version of the charm and now implemented new relation and when I upgrade the charm it seems like the charm does not recognize that it has should have relation [12:52] cnf: seems so! [12:54] anrah_: You cannot do a add-relation? [12:55] No, it says No relations found [13:02] ok, i think i have stuff sorted, now [13:02] made a special juju user for maas [13:02] now to figure out how to get a _working_ openstack deployed with juju [13:02] anrah_: can you double check that you have updated both the metadata.yaml and the layer.yaml with the interfaces of the relation? [13:03] yes, when I deploy from scratch the relation is available [13:03] But when upgrading the charm the new relation is not available [13:03] I see so it only through the update that the relation is not available [13:04] Yes [13:04] can you please open a bug or send an email to the list so that you raise awareness? [13:05] Sure! [13:08] Hello, I'm looking for a good example charm for interfacing against "postgresql" (interface:pgsql). I'm about to create my first charm with juju and have read up alot but still find it a bit daunting to create and use juju with my application. [13:09] Any help apprechiaed [13:15] hi erik_lonroth here is the list of charms using the pgsql relation https://jujucharms.com/requires/pgsql . The postgresql client seems a good place to start: https://jujucharms.com/u/postgresql-charmers/postgresql-client/3 [13:28] kjackal: Thanx! [13:30] Hey rick_h where was this zeppelin based dashboard you have? I want to steal some metrics from there:) === petevg_afk is now known as petevg [13:52] how can I get IP addresses of all application units? any code I could look at [13:54] hmm, probably relation_ids with correct relation [13:59] kklimonda: https://github.com/juju-solutions/interface-zookeeper-quorum/blob/master/peers.py#L41 [13:59] thanks [14:33] cnf: i don't know whose balls you had to scratch to get this much help from this channel but bravo [14:37] disposable2: just persistence, i guess :P [14:54] guys does anyone knows how do I connect to a mysql server that juju installed as admin user? [14:55] I do can log using the credentials of the services (like cinder nova), but I need elevated permissions [14:56] does juju support lxd 2.10/2.11? So can i use the lxd ppa? (ppa:ubuntu-lxc/lxd-stable) [14:56] i see that conjure-up now supports it, but does that implicate that juju supports it? [15:00] BlackDex: yes, conjure-up uses juju so it supports it. [15:01] oke nice [15:03] hmm, how does machine affinity work in bundle .yml files? [15:04] i see things like to: lxd/1, but lxd on what machine? [15:04] cnf: in a bundle file that would be lxd:1 then [15:04] which means lxd container on machine 1 [15:04] ah, ok [15:04] to define specific machines i suggest to create tags in maas [15:05] like machine1, machine2 etc... [15:05] can i also use space affinity do do that? [15:05] and add those as constraints [15:05] yea you can do that also. If you only want specific network space affinity [15:07] I have a small test setup, and each machine is pretty specialized, and I actually use the machine name in the tags [15:07] andrew-ii: how do you define that in the yml file? [15:08] In a bundle, I think you can put the `tags=myMachineTag` in the constraints clause [15:08] the file i have doesn't have constraints [15:08] i'll have to look up how to do that [15:08] (or to apply a bundle.yml file, for that matter) [15:09] machines: '0': series: xenial constraints: "tags=myMachineTag" [15:09] jamespage gave me an example over at #openstack-charms i'm adopting for my needs [15:09] I think it's like that [15:09] ah, hmm [15:09] interesting [15:10] hmm, so i could make tags like "controller" and "computer" and "storage" [15:10] hmm [15:10] I think so. I did that. Then realized that my machines are a bit junk and have to manually deploy specifically for the test [15:10] :P [15:11] the lightest machine is a 16 core, 24G ram gen7 HP with 500G storage [15:11] so i'll be fine, i think [15:11] nice [15:11] got 2 32 core, 96G ram machines, and a 64 core 192 G ram machine as well, for compute [15:12] That'll do [15:12] and 3 ceph nodes with a ton of disks coming in [15:12] i want to put the network gateway, and dbase and dashboard on the lightest machine [15:14] I'm currently just trying to figure out how to deploy the components without them getting hung. I keep getting incomplete relations, though I also have a mess of VLANs and spaces likely messing it up [15:14] yeah... [15:14] it's a bit of a learning curve [15:15] thankfully, $currentclient is willing to pay my hours for it :P [15:21] so i can set to: lxd right? without specifying a machine after it? or does it need a machine right there? [15:21] I think you specify the machine [15:22] But the machine may be anonymous (unless your constraints limit it to one) [15:22] hmm, you can use to: to units as well [15:22] it seems [15:23] to: lxd:neutron-gateway/0 [15:24] I think that is pretty great, if only for planning. It makes it much clearer what machine is getting what grouped with it [15:24] but form what i can see no "put it in an lxd container on a machine matching these constraints" [15:24] yeah [15:24] The constraint matching happens in the machines: clause [15:25] From then on the machine numbers are consistent [15:25] you can match constraints on application level as well [15:25] https://jujucharms.com/docs/stable/charms-bundles [15:26] I forgot about that, but I'm not sure "tags=maasTag" works for applications (though I need to try it to be sure) [15:36] cnf: You can use the constraints for lxd placements [15:36] if you use a constraint it will ask maas for a machine with that constraint [15:37] BlackDex: ok, so how do i tell it to put it in LXD, and add constraints? [15:37] from what i can see, the to:lxd expects something behind lxd [15:37] either a machine, or a unit [15:38] a machine in the case of juju [15:38] uhm [15:38] i don't follow [15:39] you define machines by number, and allocate them using constraints. After that you can use lxd: to place that specific application on that machine on an lxd [15:39] yes, i know [15:39] that's not what i was asking, though :P [15:39] ow :p [15:39] i want to know if i can put it in an lxd container. and NOT define a machine, but have a constraint instead [15:40] cnf: we don't do scheduling to lxd containers. I believe if you use just "lxd:" we'll let you, but we'll allocate a new top level machine and then put it into a container on that machine. [15:41] jam: right [15:41] jam: that's too bad, but i can work with it === hml_ is now known as hml [16:06] mwenning, yes [16:10] so lazyPwr, looks like the issue with logs might be something cert related. once I downloaded the newer 1.5.3 client I got: "Error from server: Get https://ip-172-31-12-41:10250/containerLogs/default/default-http-backend-9wcl3/default-http-backend: x509: certificate signed by unknown authority" === hml_ is now known as hml === hml_ is now known as hml [16:43] lazyPwr: around? [16:47] he has been unresponsive so far Zic [16:47] not important, thanks :) === hml_ is now known as hml [17:20] can tags be seperated by comma? [17:26] ybaumy: is this what you are looking for: https://docs.ubuntu.com/maas/2.1/en/manage-cli-tags [17:27] stormmore: Zic: lazyPwr is out sick today === frankban is now known as frankban|afk [17:30] ugh! somehow I am not surprised tvansteenburgh with how his schedule has been lately [17:31] catbus1: thanks yes thats what i was looking for. and i remember that i even saw this page before. but couldnt recall it [17:43] Hi. I'm looking for information about Juju adoption. Are there any data, any charts, or anything else one could see/use? === rogpeppe1 is now known as rogpeppe [17:59] otis you mean like a powerpoint presentation? [17:59] or documentation? [18:05] ybaumy - anything that shows any numbers and, ideally, trends [18:05] I'm asking because I've recently been asking people about Juju and most of them have either not heard of it or are not using it. [18:06] And I've been aware of Juju for *years*. So now I'm wondering if Juju is stagnant or growing or .... [18:10] hmm better ask the devs then. maybe they have something. would be interessting for me too though. if i can show management something. they think ubuntu is a kiddies OS [18:10] and juju too [18:10] alone the name they laughed about [18:11] I am surprised that I don't see it on job descriptions for sure [18:12] in the last 12 month or so. when i was getting job offers left and right. i only encountered ubuntu in a job offer once [18:12] im from germnay [18:12] germany [18:12] maybe in the US its more popular [18:14] Ubuntu is popular. But Juju != Ubuntu :) [18:14] Juju is NOT popular from what I can tell. [18:14] true but ubuntu is pushing it [18:14] ive never seen it yet on the redhat or suse sites [18:15] maybe they adopted it too [18:16] in germany everythings suse pretty much. a little redhat [18:27] in my experience there is a lot of use of Ubuntu or Ubuntu based distros in the US but not juju by any stretch [18:28] Sounds like there is nobody from Ubuntu here with numbers :( [19:36] i use juju a lot [19:36] to deploy OpenStack mainly [19:36] But also CEPH Clusters [19:37] But, i'm a ubuntu (debian) fanboy ;) [19:39] hello jujuers. Has anybody seen insert juju.system.indexes keyUpdates:0 exception: BSONObj size: 61366220 (0xCC5FA803) is invalid. Size must be between 0 and 16793600(16MB) First element: : ?type=92 code:10334 locks(micros) w:10805 5ms from juju's mongodb? [20:11] the gui is deployed to the controller right? how do i access the gui? [20:12] if i try to connect to one of the ips in the show-controller output nothing is thre [20:12] there [20:14] ybaumy: `juju gui` [20:14] ahh [20:14] nice [20:15] i seee [20:15] this is cool [21:01] I'm experiencing some very odd behavior it seems when I deploy an application multiple times in a model, almost like I'm not getting the latest version of the charm on subsequent deploys .... I don't seem to get an actual current revision that I deploy until I modify the application name [21:01] geh ... that was tattered [21:02] basically, my deploys seem to be deploying an older an older rev of the charm, even though I am specifying the latest rev when redeploying [21:02] butchered again [21:04] moreover, my lxd local deploys are acting strangely different then my lxd deploys on aws [21:05] here is a lxd local deploy http://paste.ubuntu.com/24191166/ [21:05] and here is a deploy of the same charms, but to an instance on aws http://paste.ubuntu.com/24191170/ [21:06] this has been working consistently for me across both providers up until today [21:06] in the gui is it possible to add constraints like tags for maas? [22:40] When juju's provisioner complains about there being no obvious space for a container (like a maas space), is there a way I can configure that in the bundle or gui? [22:40] My controller is on 2.1.1, if that helps (or makes things worse...) [22:45] Or perhaps a better question: does `juju gui` version 2.4.3 allow space configuration constraints?