=== frankban|afk is now known as frankban === mup_ is now known as mup [14:27] I got round to putting my kube-in-a-box project on github, brings up Kubernetes with RBAC/Istio/Docker-Registry and port forwarding on a bridged network interface all inside a Vagrant VM =) Would not have been possible without the help from here =) [14:27] https://github.com/BarDweller/Kube-in-a-box === frankban is now known as frankban|afk [19:37] with conjure-up canonical-kubernetes how do I specify a VPC? [19:37] https://docs.ubuntu.com/conjure-up/2.4.0/en/usage [19:48] I tried doing the headless on that page to specify my own controller which I manually created in the VPC, but it would just error out with the strangest of errors [19:48] 2017-11-23 19:36:44,469 [ERROR] conjure-up/canonical-kubernetes - juju.py:894 - Traceback (most recent call last): [19:48] 2017-11-23 19:36:44,469 [ERROR] conjure-up/canonical-kubernetes - juju.py:894 - Traceback (most recent call last): [19:48] File "/snap/conjure-up/842/bin/juju-wait", line 7, in [19:49] from juju_wait import wait_cmd [19:49] ModuleNotFoundError: No module named 'juju_wait' [20:16] R_P_S: https://github.com/conjure-up/conjure-up/issues could you create a bug here about that pls [20:17] hum... for reference, I didn't end up writing the last bug. Upgrading with snap --edge resolved the previous one... but this time I'm using edge for both juju and conjure-up [20:19] a regular non-headless conjure-up is in progres... although strange it seems to have hung. All the instances are listed as active with a green checkmark, but the conjure-up installer has stalled at creating a relation? [20:19] R_P_S: cool ... yea, I would like to follow the progress on that ^^ though too [20:19] the bug [20:20] is there a way to just specify a VPC with conjure up though? I thought I recalled something from days ago, but it's scrolled off chat history [20:20] and the only reference I could find was that headless option, which doesn't look like what I recalled mentioned the other day. But that might have been with respect to creating a controller in a VPC as opposed to an entire spell [20:20] you can set a default vpc to use [20:21] at the controller level I think [20:21] not so sure though [20:21] well, conjure-up makes controller and env, so I don't have any chance to change anything on the controller before the environment is spun up [20:22] ahhh, I think conjure-up will let you choose an existing controller though [20:22] Please file a bug to get that added [20:23] yeah, that's when I encountered that juju_wait error... the options for selecting a controller required also doing a headless install for some reason [20:28] damn, this vanilla "conjure-up canonical-kubernetes" has definitely hung... but I need to kill it to repro the headless errors [20:44] https://github.com/conjure-up/conjure-up/issues/1241 [20:46] that's cool... now I've not used github for submitting bugs, should I be adding a +1 comment to that FR? [20:48] R_P_S: yes, you can click the "Watch" button to be notified of activity and progress on the issues [20:49] and yes, add your 2 cents to things you care about [20:49] to a reasonable degree ;) [20:50] ok, I'm going through a complete repro of that juju_wait issue right now... rebuilding a new instance now [21:21] never used sosreport before... it just said "no valid plugins were enabled" after giving out the version (3.4) [21:21] is that expected? [21:30] https://github.com/conjure-up/conjure-up/issues/1242 [21:37] the wording on this page is a bit confusing. in the section customizing deployment phases [21:37] https://docs.ubuntu.com/conjure-up/2.4.0/en/usage [21:38] it initially talks about "controller or model" but then when it actually refers to the parameters, it calls them "datacenter" and "deployment". is datacenter=controller and model=deployment? This was the base assumption I made when I tried to run with an existing controller [21:45] Yes [21:46] It's sudo sosreport [21:47] but those docs seem to indicate that issue 1241 technically has a solution already (even if just headless), if I could just get past the bug I wrote up. Where I can create my own controller, possibly even model (with basic VPC configs for services) [21:47] that doesn't quite deal with spaces though, cause I'd still need some way to associate spaces with the apps defined inside the spell [21:47] oh ffs lol sudo sosreport [21:47] I've already downed the machine... is sosreport critical? Or should I repro again? [21:49] Nah you got the logs attached [22:25] I just reproduced that bug I wrote up with a vanilla $ conjure-up canonical-kubernetes (options: deis selected, aws credentials are full admin in IAM, set masters to 3) [22:51] R_P_S: from the looks of it, you found a legit bug [22:51] "found" :P [22:53] R_P_S: you should be excited that you helped make the software better for future use and for others by creating an issue and getting it fixed [22:53] that can be your holiday gift [22:53] :) [22:54] hehe... I'm glad to help... I just have no reason to take credit for stumbling across such a basic use case [22:55] its what you do about it that matters [22:57] to be fair, if it wasn't for the support I've received on this channel for juju, I would have abandoned it long ago and tried another option since there are so many out there [22:58] really? If you dont mind me asking, why so? [22:59] and ... becareful what you say [22:59] there is nothing like juju out there [22:59] bdx: "Please tell me honestly, but don't be totally honest"? [23:00] haha [23:00] nah [23:00] from what I can tell, I've yet to actually get a fully operational cluster online. working with my coworker yesterday, I took what I thought was a fully working cluster and found out it hadn't properly integrated with aws to create ELBs [23:00] it's become so black box that when something goes wrong, you aren't even sure what has gone wrong exactly [23:01] I've had multiple clusters complete setup without error, and yet when it comes time to do something inside kubernetes, something just didn't work [23:01] ahh, so yeah, I would say that you may have jumped in the deep end a bit fast [23:02] there's a very large gap between "quickstart" and "what I actually need" [23:02] totally [23:03] well deploying clouds and cloud software is no simple task [23:03] and the fact that I have to technically abandon conjure-up itself to eventually make what I need (aka create my own controller, model, config model, create bundle and juju deploy) [23:03] and then still not know if I'm missing something that conjure-up silently put in that I don't even know about [23:03] right right [23:03] totally [23:04] R_P_S: I feel you on all of that [23:04] right now, I'm fairly certain there are "steps" that need to be done post juju deploy bundle.yaml. But I don't know exactly how their called. I'm fairly certain one of those steps is AWS CNI. I have a hunch that was the missing piece for why all the demos always got stuck at "create load balancer with indefinite pending" [23:05] to be honest, I went with juju because it was made by ubuntu, which meant hopefully there was more than a couple guys working out of their basement. [23:06] R_P_S: thats a good reason, but there are many others that mihght make you appreciate it more if you were wise to them [23:06] but thats just the name of the game [23:06] I see kubernetes.io -> install guide -> choose one of these 75 tools. The fact alone that there's 75 tools tells me that not one of them really stands out enough to deprecate the others. [23:06] this is super cutting edge stuff, and sometime you just gotta make it work [23:06] which comes down to creating a bug and getting something fixed [23:07] I know it seems like masssive roadblock when someone is giving you a timeline [23:07] but to be fair, what you've been asked to do, with what little prerequisite knowledge you have is entirely unreasonable in the first place [23:07] which leads us to juju [23:07] well, this can start getting philosophical on the concept of MVP, Agile and and cutting edge. Cutting edge always means "replaced before it ever reaches maturity", a self fulfilling prophecy [23:08] that exists for ^ purpose kinda [23:08] haha - 75 tools; nice [23:09] which is the reason so far that I haven't abadoned juju... cause the grass is just as... brown? over there as it is over here. although maybe the grass over there was just seeded like over here [23:10] blahdeblah: 75 is a bit of an exxageration. but the table at the bottom of this "very-early" page is massive: https://kubernetes.io/docs/setup/pick-right-solution/ [23:10] and that doesn't even cover all the random solutions you'll find on github where some guy wrote it solo [23:11] the truth is, is that kubernetes is extremely powerful, and I don't honestly trust any of these solutions to really "get it right" [23:11] Its all about the operator [23:11] the solution is just a tool [23:11] please tell that to my manager [23:12] email? [23:12] prior to me joining this company, they didn't have any devops, and dev just pushed whatever they could get out. is it up? good, push it to prod. oh, it fell over? weird. [23:12] I'll kindly write him a nice letter telling him how ignorant he is [23:12] that will go over real well [23:13] jp :p [23:13] but really [23:14] its my biggest pet-peeve when people of that nature make it into positions of power in the technology sector [23:14] it hurts everyone [23:14] well, overall he's a decent guy. He's actually fairly good at treating people like humans [23:14] they usually are [23:15] but when it comes to the marketing, he just doesn't know, and has always had people around him who haven't questioned things hard enough [23:15] yeah [23:17] marketing: multicloud! what it means. most setups can be done in the cloud you're currently using. what they think: fuck yeah, we can have a cluster that spans azure, aws, inhouse baremetal and godaddy [23:17] if I was in your situation, I would start scheduling weekly educational meetings on infrastructure lifecycle [23:17] the only way is to educate them [23:17] I've been doing that for most of the year now. I've had full on training sessions about barebones infrastructure best practices, flows and pitfalls [23:18] they still have no idea what good actually looks like though [23:19] but at the end of the day, they've created a new feature, and it needs prod infrastructure yesterday cause sales is doing a demo 2 days ago to customers who have already paid for it [23:20] even my direct manager understands that there needs to be an entire culture shift from top to bottom to give us time to work out some of these underlying issues [23:20] yeah, your dead on there [23:20] Im out for t-day [23:20] roger, thanks [23:21] Im sure your conjure-up issue will be solved fairly quick, looks like an easy fix [23:21] then you can get your way [23:21] as in, I should be able to refresh --edge soon? [23:21] well, the bug you created will get status updated as the fix is released [23:21] of course, I'll watch that [23:22] I wouldnt expect anything before monday, but im sure sometime next week it will get sorted [23:32] Yes we move pretty fast .. but you know it's Thanksgiving and all [23:34] heh, I sent a video to a cubemate today. an espn announcer saying. No Games in the NHL cause it's thanksgiving, or as we call it in canada, thursday [23:35] it's from years ago, but still good for a chuckle