[00:01] babbageclunk: sure. SO for now i will address the systemd unit tests. Becuase if the fix around mongo.go is not justfied. then there is no point me fixing those mongo_tests. But i am sure the fix around that verification is ok. [00:06] vino: makes sense [00:08] I will also add a note around there in PR so that anyone looks into it can have a say abt that part in mongo.go. Maybe Ian can look into it later. [00:54] vino: can we chat in a hangout? I've got some questions about the PR. [00:55] babbageclunk: hi can u gimme 5 mins please... i am almost done fixing all issues in systemd. [00:55] please :) [00:56] or 2 mins [00:56] vino: ok, ping when you're ready. :) [00:56] babbageclunk: thank u [01:00] babbageclunk: lets go to hangout [01:00] ok [01:12] Could I get a review on this quick one? (removing debugging logging) https://github.com/juju/juju/pull/8673 [01:12] * anastasiamac looking [01:14] cheers anastasiamac [01:14] \o/ [03:01] argh, the failure reason: I'm a bid dumb doo-doo head :-\ [03:01] replace "return modelcmd.Wrap(cmd)" with "return cmd", *should* have been: return modelcmd.WrapController(cmd) [04:38] kelvinliu: FYI https://github.com/juju/juju/pull/8683 [04:38] veebers, looking now [04:38] babbageclunk: ^^ thats the PR for the work I spoke with you about the other day (add-k8s command being a BaseCommand or controller command vs a model command) [04:45] * thumper EODs [04:57] veebers: consarn it, why didn't I get a notification about that! Sorry. Looking now anyway. [04:59] veebers: approved [04:59] yay cheers === frankban|afk is now known as frankban [08:10] hi guys, I have added 10 machines using juju add-machine. if I do juju machines I get all machines 0-9 with state started [08:10] I am trying to deploy openstack-base using juju deploy ./openstack-base [08:11] but I have removed the "machines" section from the bundle.yaml file [08:11] https://paste.ubuntu.com/p/cqfrKJQkPW/ [08:12] These are the errors I am getting [08:12] if I keep the machines section, then the error is like No available machine matches constraints: [('zone', ['default']), ('agent_name', ['ca16835a-e545-49b1-89ac-d59fea465599'])] (resolved to "zone=defau [08:12] can someone help? [08:14] srihas: If you're trying to map existing machines then you might want to look into mapping them: https://jujucharms.com/docs/2.3/charms-bundles#recycling-machines [08:16] zeestrat: saved my day [08:16] thank you :) [08:18] srihas: No problemo. [08:20] zeestrat: the bundle doesn't have corosync+pacemaket thing for HA, can I add it later and map using relations? [08:20] pacemaker* [08:22] srihas: We usually deploy it straight away, but should probably work later to as long as you set the correct options. See https://jujucharms.com/hacluster/ for more details [08:24] zeestrat: will look at it, I might need to tweak a little bit :) ty [08:29] zeestrat: it installed mysql 5.6 along with the bundle not mysql 5.7 [08:29] I will wait and see if it upgrades after deployment [13:12] cinder/3 waiting idle 4/lxd/0 10.80.22.3 8776/tcp Incomplete relations: messaging [13:12] how can I see the logs? [13:12] rabbitmq-server/5 active idle 6/lxd/5 10.80.22.36 5672/tcp Unit is ready and clustered [13:14] Am I missing something? [14:30] tinwood: I updated https://github.com/juju/charm-tools/pull/396 per your review, when you have a chance to take a look. I'm also looking over your PR and I'm just about ready to +1 it. Should we consider updating the snapcraft.yaml to use Py3, though? [14:31] cory_fu, hi! Sure, I'll take at look at 396. I hadn't thought of the snapcraft.yaml to be honest :) Yes, it's probably a very good idea. Shall I do that too as part of the PR? [14:31] tinwood: Sure. [14:32] cory_fu, kk, I'll take a look at that. [14:49] o/ [14:51] with my juju deployment via maas, I can account for 17 physical servers, but inside of maas, I only see 9 nodes accounted for. 2 of the 17 are dedicated for a physical instance of vmware, 1 for an haproxy instance, so thats 3 of 17, plus 9, for 12. 5 are unaccounted for, but 89 devices are listed in maas. What are some suggestions on best ways to approach this situation? Also, I would like to move machines and applications between models [14:53] guild cats: https://github.com/juju/juju/pull/8686 [15:21] with my juju deployment via maas, I can account for 17 physical servers, but inside of maas, I only see 9 nodes accounted for. 2 of the 17 are dedicated for a physical instance of vmware, 1 for an haproxy instance, so thats 3 of 17, plus 9, for 12. 5 are unaccounted for, but 89 devices are listed in maas. What are some suggestions on best ways to approach this situation? Also, I would like to move machines and applications between models [15:25] bobeo: I would back wayyyyy out [15:25] bobeo: 1) get everything working/showing up correctly in maas [15:26] without juju being a part of the picture [15:27] bobeo: 2) bootstrap maas with juju only after you have #1 ironed out [15:28] bobeo: that should get you moving in the right direction [15:28] bobeo: so MAAS will search the network for devices out of the box on the network [15:28] bobeo: this could be wifi access points, tvs, really anything [15:28] bobeo: so the nodes view is only comissioned hardware that Juju would use [15:29] bobeo: anything on the devices list aren't in Juju's realm. That's a MAAS/network thing [15:29] if I'm reading that right [15:35] ahh I think I may have read it wrong, @bobeo my bad - sounds like your maas may be working legitimately and you are just confused about the extra devices showing up in maas? [15:35] "with my juju deployment via maas, I can account for 17 physical servers" - possibly some of these are lxd [15:35] ? [15:35] rick_h wins [15:35] :) [15:41] * rick_h_ does victory dance...if he's right [15:41] mainly looking for opportunities to dance [16:07] rick_h_ I hope your victory dance isn't one that makes it rain x) [16:09] Does anyone see why I get a TypeError method() takes 1 positional argument but 2 were given on https://github.com/Ciberth/generic-database/blob/master/reactive/generic-database.py#L35 I'm confused as I used the pgsql interface before and it worked like this :s [16:10] https://interface-pgsql.readthedocs.io/en/stable/requires.html#example-usage <-- exactly the same no? [16:19] TheAbsentOne: I'm 0/1 today so no promises [16:20] TheAbsentOne: they aren't the same, I can spot the difference, can you? [16:22] well I feel more and more retarded every day I work with the reactive framework so go ahead become 1/2 haha [16:24] TheAbsentOne: the new endpoints stuff should help with this issue because you can forego passing the arguments into the function at all and just use endpoint = endpoint_from_flag('psqldb.master.available'); endpoint.master['dbname'] - basically you are passing two objects into that function because of the two flags in the @when, and you only defin [16:24] e one argument [16:25] TheAbsentOne: https://github.com/jamesbeedy/mysql-shared-example/blob/master/reactive/mysql_shared_example.py#L39 [16:26] that's what I was thinking the second flag passes a second object so basicly 2 'selfs' right? bdx [16:26] so in this example I can just delete the parameter then? [16:26] bdx: rick_h_ Yea, its working right, and synced right with juju. my issue is I have hardware that is an outlier to juju/maas, and I want to keep it that way for that existing infrastructure. I made the mistake of not configuring the IPMI intelligently, which was my mistake, and so now Im trying to unfubar myself [16:27] lol ok then :) [16:27] and what is the proper way to have the object then where I can call set_database on in the function above bdx ? Can you do that too with endpoint? [16:27] The idea being identify the other 8, which I have now,a nd figure out A. How did I miss 8 servers? and B. Reconfigure their IPMI networking so they arent mixed in, so I dont make the same mistake twice. [16:27] TheAbsentOne: yes [16:28] TheAbsentOne: another example https://github.com/omnivector-solutions/layer-logstash/blob/master/reactive/logstash.py#L53,L56 === frankban is now known as frankban|afk [16:28] so i can endpoint = endpoint_from_flag('..'); endpoint.set_database()? bdx [16:28] yessir [16:29] that's awesome, thanks for that example of logstash [16:29] np [16:29] I'm really messing things up by combining new ways with old ways it seems [16:33] TheAbsentOne: yeah, things just recently got changed around a bit, I'm totally in favor of the endpoints patter (the new way) - I feel its far cleaner and more explicit amongst other things [16:33] TheAbsentOne: its probably just more confusing for you combining the two [16:36] TheAbsentOne: the olden way should be gone and deprecated soon enough, I would just stick with the endpoints pattern going forward unless you have a reason not to [16:55] yeah definitely the endpoint pattern is awesome, it's just that I try to look at other examples for help and stuff and I mix it up doing weird stuff [16:55] I'm gonna try to fix it in a bit thanks again bdx, maybe you should write a tutorial ;) [16:56] np [16:57] TheAbsentOne: seems to be all I do lately - https://imgur.com/a/lgEkHKn [16:58] thats only a small subsection of the overall docs [16:58] Hehe shows how fluent you are with it! I'll try to write some things to and let you guys review it, might be interesting for newcomers [16:58] is it public available? bdx [16:59] TheAbsentOne: awesome [16:59] not yet [16:59] they will be soon [16:59] looking forward to it! [16:59] getting lots of good feedback internally, should give for a better end product when I release them to the wild [17:00] give me a week or so, I'll ping the list when I start releasing them [17:02] Good stuff, sounds very nice [17:22] Juju Show reminder in 40min kwmonroe cory_fu bdx zeestrat and anyone else that wants to hang out or stream [17:22] get [17:22] get your coffee filled up [17:41] ahoy rick_h_, i won't be able to make it today -- we welcomed a baby girl into the world last week, so i'm just getting caught up from paternity leave. i'll see your coffee advice and raise you a redbull. [17:41] kwmonroe: congrats!!!!! [17:41] kwmonroe: time to setup the mainline [17:41] :) [17:41] kwmonroe: assume everyone is doing ok? [17:42] yup yup rick_h_, everyone is doing great! [17:43] kwmonroe congratz, enjoy it! [17:43] thanks! [17:43] kwmonroe: so awesome, get a good nap in :) [17:51] bdx: rick_h_ ok so I solved that problem, now juju is trying to deploy in two zones I dont want it to deploy from, which arent the defalt zone. any ideas? [17:52] Links for the show today: to join and chat https://goo.gl/4kZPBT and to watch all the things from the comfort of an un-mike'd browser use https://www.youtube.com/watch?v=HIDcHiNV-6o [17:54] I still love that foto and the fact it was taken at my college x) [17:54] TheAbsentOne: cool [17:55] kwmonroe: congrats!! [17:55] thx bdx! [18:11] rick_h_ bdx kwmonroe how do I deploy to a specific zone? [18:12] bobeo: --to zone= [18:13] bdx: HAZAAAH! [18:13] I kept doing --zone= I thought I was going crazy XD [18:15] bdx: tell rick_h_ that his screen is not the primary [18:16] yeah not catching the terminal [18:17] nope, bdx, tell him to screen share better -- he's showing Erik's face [18:20] bdx: ok so Im losing my mind here. Does the hardware state need to be in "Ready" or "Allocated" state in order to be allocated to a juju application? [18:20] bobeo: ready [18:23] wait I was behind a bit, what happened to the stream xD [18:23] TheAbsentOne: it died for me around 15m :/ [18:23] probably because i was lamenting the fact that rick_h_'s terminal was not being presented. [18:24] yeah same, too much eric for the stream to handle x) just kidding ;) [18:24] lol [18:24] the call is probably still busy? [18:27] uh oh [18:27] bdx: Ok, so then why isnt this working? juju deploy cs:~graylog-charmers/graylog --to zone=Security [18:27] when I dropped to reconnect did it kill the livestream? [18:27] possibly [18:27] the box is in the ready state in the Security zone [18:27] oh crap it did!?!?!?! [18:28] rick_h_: youtube posted vid of almost 16 min [18:28] oh man...rick fail [18:28] even worse than I thought I did [18:28] all is lost after that xD [18:28] bobeo: whats does the error say? [18:28] well actually it's there after I reconnect [18:28] so that didn't kill it, but something else did [18:29] do you have local recording rick_h_ ? [18:29] TheAbsentOne: nope...I had to switch machines and such [18:29] bdx: zeestrat (bdx is your friend here?) https://insights.ubuntu.com/2018/05/03/lxd-clusters-a-primer [18:29] bdx: Thats the weird part, its not giving an error. The system isnt shifting from ready to allocating, maas specifically, and juju is just saying pending machine [18:30] agent says "allocating" [18:30] should I kill the app and restart it? [18:30] idk, @eriklonroth^ [18:30] the install* [18:30] the node is probably booting [18:31] bdx: thats the thing though, its not moving from the ready position in maas, its just sittng there [18:31] bobeo: ahh I see [18:31] kwmonroe: oh damn, I see what you're saying. Since I reconnected and wasn't the "driver" of the stream I could click and see my terminal but it didn't make it the stream's main view [18:31] man, what a can of fail today [18:32] ok, pulling the video [18:32] bobeo: can you capture a few screen shots [18:33] of the node details page in maas and the juju status [18:33] sure! [18:33] rick_h_: don't be bothered by it too much man it happens [18:35] hah, /me just put together eriklonroth is erik from the HO. https://insights.ubuntu.com/2018/05/03/lxd-clusters-a-primer for you too [18:35] bdx: ok I got them [18:36] bobeo: use your fav img posting service and send some links [18:38] bdx: https://imgur.com/a/vt20E8A [18:39] bdx: Do you think its because of this from the link vs the instruction set? cs:graylog-15 vs juju deploy cs:~graylog-charmers/graylog --to zone=Security [18:40] bdx: Do you think if I changed it to juju deploy cs:graylog-15 --to zone=Security that it will deploy? [18:40] bobeo: what I usually do in this situation is go back to square 0 [18:40] bobeo: try just `juju deploy ubuntu` [18:40] or `juju add-machine` [18:41] we need to get the most basic use case working before trying to deploy charms to zones [18:41] possibly `juju add-machine` may be best here [18:42] see if you can get that to work [18:42] bdx: see if I can get a machine to add at all, then add the app to the machine with the --to machine:"X" sufix? [18:43] yeah, but more or less, we just want to verify [18:43] this isn't the long term fix [18:43] dangit, its hung not aborting two add-machine commands [18:43] tried juju remove-machine "X" and with --force, its not working [18:43] after you verify that you can add machines successfully, then try adding them to specific zones `juju add-machine zone=Security` [18:43] still trying to add the machines [18:44] yea, thats what I meant to do [18:44] I see [18:44] bobeo: yeah, you may have gotten your model stuck possibly [18:48] bdx: Any ideas? I checked all IPMI to make sure no mismatched IP, and no issues there. [19:57] bdx: you still here? I'm close to having a working thing but I'm once again lost in my own thoughts. I basicly want to pass a flag from one charm to another. How do you that in endpoint pattern properly? [20:00] I want a @when in my other charm on this https://github.com/Ciberth/generic-database/blob/master/reactive/generic-database.py#L70 [20:04] https://www.dropbox.com/s/n27pznx2ms3ew4d/temphandlerquestion.png?dl=0 <-- this in the requires.py doesn't make sense right? [20:37] Morning all [21:03] morning team [21:20] morning [21:32] Welp I see the morning's time for me to go to bed, I'll try again tomorrow ^^ bye everyone [21:41] so... if you trash your laptop [21:41] is it possible to reconnect to a manually provisioned juju setup? [21:41] or do i need to find a backup from somewhere? [21:43] magicaltrout: was the controller on your laptop too? [21:43] nope [21:45] i tried `juju register ` [21:45] magicaltrout: you could run 'juju add-user' on the controller host [21:45] I have zero clue if its doing anything [21:45] magicaltrout: that shuld probably work but i guess you still need some creds - maybe try with --debug [21:46] debug shows nowt also [21:46] magicaltrout: try 'juju add-user' on the controller machine [21:47] hmm so the controller is remote, but I didn't bootstrap it on the remote machine [21:47] so there's no juju binary [21:48] hrm [21:48] magicaltrout: I feel like this must have happened to someone before... [21:48] meh one of the other guys has access to it I can get him to prod it and sort it out tomorrow, i just trying to cheat [21:49] I think you could make the controllers.yaml file you'd need, given that you can ssh to the controller machine, but it would be fiddly. [21:51] yeah.... not that bothered ;) I'll figure it out via a working install tomorrow :P [21:51] thanks chaps [21:53] \o/ [21:54] magicaltrout: cool [22:02] magicaltrout: I strongly recommend keeping a backup [22:03] backups are for wimps [22:03] magicaltrout: in theory it is possible to reconstruct the components necessary to build up what you need to reconnect [22:03] but it isn't easy [22:03] and not automated [22:04] if you had credentials on the controller [22:04] then IIRC there was a connect command [22:04] either done or planned [22:04] * thumper looks [22:05] well this doesn't make sense at all, my initial bisect suggests that 9d4339d (inc to beta2) is good and 9d4339d is the first bad (adding caas unit test bits). How can that break the CMR test 0_0 [22:07] magicaltrout: if the controller doesn't have a public cert, there is no easy way to reconnect [22:07] as the client doesn't know the server CA cert [22:07] so there is no client command to do this [22:08] who cares about SSL, GCHQ can crack it anyway..... [22:08] fair enough thumper ! [22:08] np [22:28] babbageclunk: got a few minutes? [22:28] thumper: yup - give me 1 min? Just finishing a review. [22:28] ack [22:32] thumper: ok [22:32] in 1:1? [22:32] yeah