[10:26] <gnuoy> Can I specify a relative local directory as the charm source in a bundle ?
[10:27] <admcleod_> gnuoy: pretty sure you can
[10:29] <gnuoy> Everything I try results in one form of explosion or another eg https://paste.ubuntu.com/p/rdFhVWn6kg/
[10:29] <gnuoy> Fully qualified paths seem fine
[10:34] <admcleod_> gnuoy: try specifying the series as an option
[10:35] <admcleod_> gnuoy: for the vault charm
[10:36] <gnuoy> admcleod_, hurray \o/
[10:36] <gnuoy> admcleod_, thank you
[10:36] <admcleod_> gnuoy: it looks like it assumes a series, so charm: xenial/vault would work if you also had ./xenial/vault
[10:37] <gnuoy> that didn't seem to work
[10:37] <admcleod_> gnuoy: i know ive done something like that in the past but perhaps its changed
[10:37] <admcleod_> gnuoy: i guess series specification isnt a huge issue
[10:37] <gnuoy> I mean xenial/vault didn't, the series option is golden
[10:37] <gnuoy> admcleod_, not to me, I'm happy to restate it
[10:55] <TheAbsentOne> How can one use a sql file in a hook to create and populate a database with the reactive framework properly? Or is there an option in the mysql charm itself for this?
[11:17] <rick_h_> TheAbsentOne: so normally there'd be some flag set that the db is up and ready and at that point in time you'd use whatever db library to connect with the information stored from the relation and such and write out the data to the database.
[11:19] <TheAbsentOne> yeah got to that point, I'm just not that familiar with how python can achieve this properly but Im testing things out right now, thanks rick_h_
[11:19] <rick_h_> TheAbsentOne: what db is it?
[11:22] <TheAbsentOne> the app will use and need both a mysql and mongo connection but for now I'm focussing on mysql
[11:26] <rick_h_> TheAbsentOne: check out https://github.com/PyMySQL/PyMySQL
[11:26] <rick_h_> TheAbsentOne: you should be able to add it to the requirements file for the charm I think and then use that from within the hooks
[11:27] <TheAbsentOne> ohn nice that should work rick_h_ thanks for leading me somewhere! ^^
[11:28] <rick_h_> TheAbsentOne: no problem, have fun!
[12:59] <TheAbsentOne> rick_h_: still here by any chance? If 1 charm needs both a mongodb and a mysql charm for example can juju handle this like the first method or doe I need the second way (https://pastebin.com/0GbiHFE4)
[13:00] <magicaltrout> you'd need the second TheAbsentOne
[13:01] <TheAbsentOne> I thought so, and can the charms from the charmstore handle this? magicaltrout
[13:02] <magicaltrout> i'm not really following what went before.... handle what?
[13:03] <magicaltrout> i'm working on a jdbc relation for example that wouild sorta evenually do what your first example did, but to my knowledge, if you want mysql and mongo related in one charm, that'll be fine as long as you have 2 hooks and tell them what to do
[13:03] <TheAbsentOne> there's nothing special actually. (It's for a POC) I'm gonna use the apache php layer and I have a few php files I want to deploy as a service. One of these files reads from a mongodb and anoher from a mysql one
[13:03] <magicaltrout> yeah thats fine, you'll have 2 functions one for mysql available and the other for mongo available
[13:04] <TheAbsentOne> are you really working on that? That's basicly one of the subjects of my dissertation here! :o
[13:04] <magicaltrout> well actually you could nest it in one probably
[13:04] <magicaltrout> buit from a usability point of view you'd probably want to feed back to the user if one service wasn't available and the other was
[13:04] <TheAbsentOne> yep that was my first step and in a next iteration I kinda want to build a "generic type of database"-charm
[13:04] <TheAbsentOne> correct!
[13:05] <magicaltrout> i've got an apache drill charm so i've got a jdbc relation
[13:05] <magicaltrout> thats generic enough to do stuff
[13:05] <magicaltrout> and then my ui charm can accept that
[13:05] <magicaltrout> from mysql, drill, postgres or whatever
[13:05] <TheAbsentOne> is it on github? I might learn a stuff or two
[13:05] <magicaltrout> although i've not upstreamed mysql support yet
[13:05] <magicaltrout> maybe let
[13:05] <magicaltrout> me look
[13:06] <TheAbsentOne> if you are interested btw, I did some basic writing on a gitbook about my research: https://ciberth.gitbooks.io/logboek-thesis/content/problem.html
[13:09] <magicaltrout> https://gitlab.com/spiculedata/juju/interface-jdbc
[13:09] <magicaltrout> doesn't do much currently other than offer up some jdbc details
[13:09] <magicaltrout> borrowed mostly from the mysql interface
[13:12] <TheAbsentOne> still usefull! Thanks a lot
[13:12] <TheAbsentOne> It's pretty hard finding my way through all the documentation and examples
[13:12] <TheAbsentOne> as a novice
[13:13] <magicaltrout> tell me about it
[13:14] <magicaltrout> the only way i understood anything was from sitting with kwmonroe and cory_fu for about 6 hours in Gent
[13:14] <magicaltrout> turns out kwmonroe hacks for beer
[13:14] <TheAbsentOne> It's actually funny how we have this big paradox. We want to simplify things by sometimes adding complexity
[13:15] <TheAbsentOne> in Ghent? You are from Belgium? :)
[13:15] <magicaltrout> nope
[13:15] <magicaltrout> but if someone pays
[13:15] <magicaltrout> i'll go anywhere
[13:15] <TheAbsentOne> hehe as a matter of fact I'm from Ghent city xD
[13:15] <magicaltrout> there was a juju track at devops days twice
[13:15] <magicaltrout> so we all met up there
[13:15] <TheAbsentOne> and at configmngmtncmp a few years back, that's where I got to know juju
[13:16] <magicaltrout> yeah
[13:16] <magicaltrout> i was lying you got the correct name :P
[13:16] <magicaltrout> i will have been lurking at the back somewhere
[13:18] <TheAbsentOne> hehe
[14:18] <TheAbsentOne> what can cause an unknown status and workloud even if every function has a status_set: https://github.com/Ciberth/mycustomcharms/blob/master/firstiterations/step0/reactive/step0.py
[14:19] <TheAbsentOne> loggin in with ssh shows that he didn't even install apache and debug hooks doesn't really work, as there is no hooks folder or anything yet
[14:20] <rick_h_> TheAbsentOne: that means you forgot to run charm build and to use the built charm to deploy
[14:20] <rick_h_> TheAbsentOne: a common error that there's definitely a bug about warning a user when no hooks found
[14:21] <rick_h_> TheAbsentOne: have to get into the habit of checking charm build and charm proof to sanity check before deploying/upgrading/testing
[14:22] <TheAbsentOne> oh god you are right I build it but I still ran the well not the builded one xD guess I'm tired. Thanks rick_h_ I could have lost a lot of stupid time here
[14:22] <TheAbsentOne> charm proof said everything was fine xD
[14:22] <rick_h_> TheAbsentOne: yea, I usually end up with one terminal in the charm code directory and one in the build directory so I split the work of editing/etc of the charm and deploy/juju interactions while testing
[14:23] <TheAbsentOne> that's exactly what I'm gonna do, I feel so stupid lol
[14:23] <magicaltrout> i've done that about 6000 times
[14:23] <rick_h_> don't, you don't know how many times I've done that
[14:24] <rick_h_> well...not sure it's 6k but magicaltrout is generally more #winning than me :P
[14:24] <TheAbsentOne> yeah kinda figures that it happened to others before
[19:59] <rick_h_> kwmonroe: around? having issues auth'ing to the k8s webui and curious if I'm missing something
[19:59] <kwmonroe> yup rick_h_
[19:59] <kwmonroe> rick_h_: do you see the login prompt on the webui, or does it show some gobbledeegook json/xml?
[20:00] <rick_h_> hey, so I have the proxy going and I go to my /ui URL and get redirected asked about uploading a config file or using a service account token
[20:00] <kwmonroe> rick_h_: SKIP
[20:00] <rick_h_> kwmonroe: so I point it at my .kube/config and it says "Authentication failed. Please try again."
[20:00] <rick_h_> oooooh...ummm...wtf?
[20:00] <kwmonroe> isn't there a skip option?  i never use that config file doohickey
[20:00] <kwmonroe> just type admin/<password> and go about your business
[20:01] <rick_h_> yea, fair enough, there is a skip. I guess I kind of read that like "GO" and "Cancel" and didn't catch it was skip
[20:01] <rick_h_> I didn't even have to do that much
[20:01] <rick_h_> it took me straight to the dash
[20:01] <kwmonroe> to be fair rick_h_, i have no idea what that token/config would allow you to do that skipping prevents.
[20:01] <rick_h_> oh hmm, but a crossed out head in the corner
[20:01] <rick_h_> if I click that it'll bring back up the dialog
[20:02] <rick_h_> kwmonroe: so where do I go to auth with user/pass?
[20:02] <rick_h_> kwmonroe: if I go to "settings" in the dashboard I get a 403
[20:02] <kwmonroe> hm, you should have been prompted for a user/pass first, then you'd get the "upload a config / token" -- the latter is what i skip
[20:02] <rick_h_> kwmonroe: nope
[20:03] <kwmonroe> rick_h_: did you conjure-up this thing, or juju deploy?
[20:03] <rick_h_> kwmonroe: juju deploy canonical-kubernetes
[20:03] <rick_h_> well, I had to download the bundle to put etcd on containers so I can fit it in 6 machines :/
[20:03] <rick_h_> but same diff
[20:05] <kwmonroe> rick_h_: does "kubectl cluster-info" give you output that says "k8s dashboard available at:  blah"?
[20:06] <rick_h_> kwmonroe: rgr it does
[20:06] <kwmonroe> and is that the URL you're trying to access?
[20:06] <rick_h_> kwmonroe: no, I ran kubeproxy and went to the 127. address
[20:06] <kwmonroe> because maybe you're trying to hit the master/ui directly versus the proxy
[20:06] <rick_h_> per the instructions :)
[20:06] <kwmonroe> oh mylanta, didn't know we had instructions
[20:06] <kwmonroe> ryebot: help this poor soul ^^
[20:07] <rick_h_> ah yea, hitting the url directly w/o proxy has a basic http dialog
[20:07] <ryebot> lookin'
[20:07] <kwmonroe> ok - try admin and the password from ~/.kube/config-$JUJU_MODEL
[20:07] <rick_h_> hah, so I can login with my .kube/config file except that it doesn't really log me in
[20:08] <rick_h_> it let's me through the htaccess stuff but I still get 403 on settings/etc
[20:08] <kwmonroe> you have to login, then click skip, then physically skip around the room, then press cancel.
[20:08] <rick_h_> kwmonroe: so I argue one cannot login to the dashboard properly
[20:08] <kwmonroe> lol
[20:08] <rick_h_> kwmonroe: ooooh, I forgot the cancel part
[20:08] <kwmonroe> your argument is noted
[20:08] <rick_h_> :P
[20:08] <ryebot> rick_h_ use the kubeconfig from /root on a k8s worker
[20:09] <rick_h_> ryebot: not the instructions "juju scp kubernetes-master/0:config ~/.kube/config" ?
[20:09] <rick_h_> that's the config I'm using, is the worker ones different?
[20:10] <ryebot> there's two sets of credentials, for whatever reason the ones in /root on a worker get me logged in
[20:10] <ryebot> they're probably somewhere on the master too, but idk where
[20:10] <rick_h_> ryebot: /root/cdk/kubeconfig ?
[20:10] <rick_h_> ryebot: there's no config in /root
[20:10] <rick_h_> ryebot: my /root looks like "cdk  curtin-install-cfg.yaml  curtin-install.log  snap"
[20:10] <ryebot> rick_h_ yeah that sounds right, sorry going off memory here
[20:11] <rick_h_> ryebot: gotcha, ok
[20:11] <rick_h_> ryebot: ah, yea the kubeconfig is a service token vs a user account
[20:11] <ryebot> it's ... a little screwy and kinda grew into that organically. I think we have a task to make that less awful.
[20:12] <rick_h_> ryebot: hmm, ok. I'm working out how things work so I'll tinker and find the service account stuff and see if I can list/create them and such then
[20:13] <rick_h_> so thanks for skip kwmonroe and ty ryebot for saying I'm not holding it wrong
[20:13] <ryebot> xD
[20:14] <kwmonroe> does that mean you're in rick_h_?
[20:16] <tvansteenburgh> we should update the instructions
[20:18] <tvansteenburgh> and we should not be telling users to auth as service accounts :P
[20:21] <tvansteenburgh> i'm trying to remember why we used to kubectl proxy to reach the dashboard
[20:21] <tvansteenburgh> i haven't done that in ages
[20:22] <tvansteenburgh> although i would expect it to work with basic auth anyway
[20:25] <knobby> tvansteenburgh: what is your preference? an ingress or loadbalancer depending on substrate?
[20:28] <tvansteenburgh> knobby: well with cdk the lb is already there
[20:30] <knobby> tvansteenburgh: so the load balancer for the api server would pull double duty to pass off to the dashboard service?
[20:31] <tvansteenburgh> eh, doesn't it already? if i kubectl cluster-info, my dash url is the lb ip address
[20:31]  * knobby isn't certain
[20:32] <knobby> I'm just trying to figure out what the goal is here
[20:33] <tvansteenburgh> knobby: the problem is that rick_h_ followed our bundle instructions, which tell you to run kubectl proxy to access the dash, but then he discovered he couldn't basic auth using that method
[20:34] <tvansteenburgh> i haven't tried this myself yet
[20:34] <tvansteenburgh> i can't remember the last time i used kubectl proxy though
[20:36] <knobby> tvansteenburgh: right, I get that the general problem is that the CDK instructions aren't as good as they should be for the dashboard and I'd like to see what we need to do to fix it. But I think step one there is to define the ideal experience. How do we talk to the dashboard? Is it an ingress, through a lb, or hidden behind kube proxy? Then we work towards that. kjackal has done some work here recently and
[20:36] <knobby> might have some opinions...
[20:37] <knobby> tvansteenburgh: I don't use my dashboard either and that hurts my perspective here
[21:09] <rick_h_> tvansteenburgh: the bigger issue is that you cannot log into the dashboard login UX with your config
[21:09] <rick_h_> tvansteenburgh: I'm not sure how that's supposed to work, but the proxy just bypasses the http basic auth and you end up at the dashboard and logging into that itself doesn't work
[21:10] <rick_h_> kwmonroe: so no, I'm into the main dashboard but as an anon user vs as "admin"
[21:10] <rick_h_> that's why things like the settings nav menu give me a 403 (I assume)
[21:12] <rick_h_> kwmonroe: tvansteenburgh let me know if you want me to screenshare what I mean if I'm not being clear