magicaltrout | if nothing else kwmonroe at least Hue shows me whats missing and whats completely borked currently in the Big Data stack :P | 00:07 |
---|---|---|
veebers | kelvin_: can you remember what the "extra thing" was to use docker installed from snap? i.e. so you didn't need to use sudo | 02:05 |
kelvin_ | veebers, the thing was snap version of docker was a little bit out of date, so the flag we need was introduced does not work | 02:07 |
veebers | kelvin_: ah, sorry this is me using it locally on my machine. I tire of having to use 'sudo', I'm sure there was something to work around that, but perhaps it required a logout or something | 02:07 |
kelvin_ | veebers, it's not related the user group. | 02:08 |
kelvin_ | veebers, if u wanna use docker without sudo, then u just need to add ur user to the docker group | 02:08 |
veebers | kelvin_: ah I see, it's in the readme https://github.com/docker/docker-snap | 02:08 |
veebers | kelvin_: also, that states that docker snaps won't be updated, so you made the right choice with the install type for that job! L:-) | 02:09 |
kelvin_ | aha :-> | 02:10 |
veebers | wallyworld: Hmm, I'm having trouble with the charm command trying to push a charm with a docker resource (I get: unsupported resource type "docker", after adding a filename to the resource as it complains about that too which it shouldn't) | 02:38 |
veebers | this is with the edge snap of charm installed | 02:38 |
* veebers tries building from source | 02:45 | |
wallyworld | veebers: is that a cmd message or one from the store? | 02:48 |
veebers | wallyworld: good question, I assumed from the command. I do have JUJU_CHARMSTORE exported, let me check I have that setup properly | 02:49 |
veebers | wallyworld: that seems fine, charm whoami stats I'm not logged into the staging | 02:50 |
veebers | ah shoot, I need to remember my username/password now ^_^ | 02:51 |
wallyworld | lol | 02:51 |
veebers | ah wait, which sso does the staging use? Probably staging sso? | 02:51 |
veebers | hmm, charm login doesn't work anyway | 03:00 |
veebers | wallyworld: yeah, if I set JUJU_CHARMSTORE to the staging env, I can't charm login. I'll fire off an email | 03:05 |
wallyworld | veebers: ah bollocks ok | 03:09 |
thumper | babbageclunk: just a quick check, but this lease refactoring keeps both bits working right? | 03:33 |
babbageclunk | thumper: both bits meaning singular and leadership? | 03:33 |
babbageclunk | yes, I think so - tests all pass. | 03:34 |
thumper | both bits meaning mongo and raft | 03:34 |
babbageclunk | raft isn't there yet | 03:34 |
babbageclunk | I mean, it's there but doesn't know about leases | 03:34 |
thumper | ok, so this is just paving the way? | 03:35 |
babbageclunk | yup | 03:35 |
babbageclunk | But the idea is that they'll both be working, and we can select between them at bootstrap time. | 03:35 |
thumper | um... | 03:35 |
thumper | at controller config time? | 03:36 |
thumper | don't we want to be able to switch in a running system? | 03:36 |
babbageclunk | Well, maybe, if we're alright with all of the leadership changing at that point. | 03:36 |
thumper | I think that'd be fine. | 03:36 |
thumper | I've been thinking a little | 03:36 |
babbageclunk | Then yeah, that should be fine | 03:36 |
thumper | it would be nice if an operator could say "make that unit the leader" | 03:37 |
thumper | just something to think about | 03:37 |
thumper | consider our rolling upgrades | 03:37 |
thumper | upgrade a single non leader unit | 03:37 |
babbageclunk | Yeah, you've mentioned something like that before. | 03:37 |
thumper | then make it the leader | 03:37 |
thumper | then upgrade the others | 03:37 |
thumper | something like that... | 03:37 |
thumper | I imaging that there will be some weird edge cases... | 03:37 |
babbageclunk | Yeah, definitely - what if that machine dies first. | 03:38 |
babbageclunk | But worth a try. I'm keen to get the existing stuff working first though. | 03:38 |
thumper | definitely | 03:38 |
kelvin_ | wallyworld, can i have ur a few minutes to dicuss charms for gpu testing? | 03:43 |
thumper | babbageclunk: fyi tests failed, didn't check what | 03:51 |
babbageclunk | thumper: ooh, thanks | 03:52 |
veebers | babbageclunk, thumper: is there a way in a JujuConnSuite (or so) to set a controller config? If so I can alter these deploy test just a little to cover using the charmstore url from controller config across the board | 05:03 |
babbageclunk | veebers: there is, I'm just trying to remember how you do it. | 05:03 |
babbageclunk | veebers: Try setting s.ControllerConfigAttrs in the SetupTest before you call s.JujuConnSuite.SetupTest | 05:05 |
veebers | babbageclunk: awesome, cheers! I'll give that a bang | 05:06 |
babbageclunk | veebers: you can see an example in apiserver/admin_test.go | 05:06 |
veebers | babbageclunk: is s.Session.DB setup by the ConnSuite? | 05:08 |
babbageclunk | uhhh | 05:08 |
veebers | I ask because I need to get the charmstore url created by new server as a controller config, if I need to do that before conn.SetupTest then I have an issue ^_^ | 05:08 |
veebers | I think it does | 05:09 |
babbageclunk | oh, yeah, it's setup in MgoSuite.SetUpTest, which gets called from JujuConnSuite.SetUpTest. | 05:11 |
babbageclunk | Sounds like you need a closed time-like loop. | 05:11 |
babbageclunk | So what generates the charmstore url? | 05:12 |
babbageclunk | veebers: ^ | 05:12 |
veebers | babbageclunk: charmstore.NewServer(db, nil, "", params, charmstore.V5) creates a handler which is then passed into httptest.NewServer(handler), the result of that has the URL. (not db in NewServer is from s.Session.DB("juju-testing")) | 05:13 |
babbageclunk | hmm | 05:13 |
veebers | babbageclunk: hah is MgoSuite.SetUpTest idempotent? ^_^ | 05:14 |
veebers | no, it's not | 05:14 |
babbageclunk | worth a try though | 05:15 |
veebers | babbageclunk: hmm, actually being able to do that might not be as useful as I originally thought, so not a biggie that I can't. Thanks for sorting me out on that though | 05:17 |
babbageclunk | You could extend the ControllerConfigAttrs handling so that if there's a callback set it calls it to get the config attrs | 05:17 |
babbageclunk | And by default the callback just returns s.ControllerConfigAttrs | 05:17 |
babbageclunk | That would give you a chance to get the url | 05:18 |
babbageclunk | It's awful but the JujuConnSuite setup is already terrible | 05:18 |
veebers | ^_^ | 05:19 |
veebers | babbageclunk: quick query, I'm doing this, is there a better way? https://pastebin.canonical.com/p/WVgsk2pKbf/ | 05:32 |
veebers | as the patched function now takes the url to use, but it needs to be that one specifically | 05:32 |
babbageclunk | looking... | 05:33 |
babbageclunk | oh, I see - yeah, I think that's ok. Not sure how else you could do it, without knowing the rest of the code. | 05:35 |
veebers | babbageclunk: ack, cheers. wasn't sure if there was a better way than taking a ref and using that | 05:35 |
babbageclunk | Well, depending on what else it does if you could bypass calling the patched out function that would probably be simpler. | 05:37 |
babbageclunk | but I think that's fine. | 05:37 |
veebers | The patched function does what was returned pretty much (at this point at least) but felt it was patching too much, would be nice to not even need the patch | 05:38 |
babbageclunk | yeah, if you can avoid it better not to, but sometimes that's really difficult. | 05:40 |
babbageclunk | ugh, just found a test that passes if it's run with the others in its suite but fails when run by itself (or in a ci build, apparently) | 05:42 |
veebers | ugh :-\ that's going to be fun to nail down | 05:43 |
babbageclunk | well, easier than the other way around | 05:44 |
veebers | true, half-full then :-) | 05:44 |
veebers | wallyworld: FYI https://github.com/juju/juju/pull/8896 | 05:51 |
veebers | ah damn, the formatting for the PR comment is borked, I misunderstood what it would look like from the vim buffer 'hub pull-request' bought up | 05:52 |
* veebers will fix that after dinner | 05:52 | |
vino | wallyworld: do u have min to discuss regarding the version increment | 06:51 |
wallyworld | sure | 06:51 |
vino | HO | 06:51 |
vino | wallyworld: sorry i was a bit away during ur discussion. | 07:14 |
wallyworld | no problem, i'll jump back in | 07:14 |
vino | :p | 07:15 |
vino | k thank u | 07:15 |
stickupkid | I'm guessing that file shouldn't be there - https://github.com/juju/juju/blob/develop/apiserver/dependencies.tsv | 10:08 |
stickupkid | PR of removal https://github.com/juju/juju/pull/8898 | 10:13 |
JaniferHe | help | 12:01 |
JaniferHe | juju status | 12:01 |
rathore_ | Anyone faced any issues with bionic host and bionic lxc containers? | 15:05 |
rathore_ | My bionic containers are started and has IP ( saw in host) but Juju is still thinks it doesnt have IP | 15:05 |
hml | rick_h_: so it turns out there is a MinRootDiskSizeGiB() used by vsphere and gce, but not the others.- 8Gb. | 15:29 |
rick_h_ | hml: ah gotcha that makes sense | 15:30 |
rathore_ | rick_h_: Would you have any idea. juju starts bionic lxc containers and wait indefinitely | 15:33 |
rick_h_ | rathore_: no, there were some issues addressed in 2.4.0 with pulling in maas resolve info and such but nothing I can think of about not getting an IP? | 15:34 |
rathore_ | rick_h_: sudo lxc list on host shows me that IPs are there. I have tried 2.4 and 2.4(beta) | 15:34 |
rick_h_ | rathore_: is there anything in the debug-log about the machines not coming up? anything more in juju status --format=yaml around the machines status? | 15:35 |
rathore_ | I am just deploying again and would paste as soon as I see any logs | 15:36 |
rathore_ | rick_h_: https://paste.ubuntu.com/p/HdDw4HDWCR/ is from yaml output | 15:46 |
rathore_ | rick_h_: Ok, so the containers are still downloading the tools from controller | 15:56 |
rathore_ | somehow the networking in container is messed up | 15:56 |
rathore_ | rick_h_: The file downloads are fine on host but toooo slow in container | 15:57 |
rathore_ | rick_h_: It seems to be lxdbr0 issue, it is not connected to any network | 16:08 |
rathore_ | https://paste.ubuntu.com/p/6YqtgNBpxZ/ | 16:08 |
cory_fu | rick_h_: I've got some updates for conjure-up and cloud integrator charms for the Juju Show, FYI | 17:30 |
rick_h_ | cory_fu: woot woot | 17:31 |
stokachu | rick_h_, is that discourse board going to be officially staying? i want to direct all non bugs from github conjure-up to that forum | 17:31 |
rick_h_ | stokachu: yes, we're just ramping it up | 17:32 |
stokachu | rick_h_, is it ok if i start pointing people there? | 17:32 |
rick_h_ | stokachu: by all means | 17:34 |
stokachu | thanks | 17:34 |
rick_h_ | stokachu: I'm going to bring it up on the show and we're slowly starting to port docs/notes/etc and will kill the lists in a few | 17:34 |
stokachu | rick_h_, ack, sounds good | 17:34 |
cory_fu | rick_h_: Oh, and if there's time for it, also a change to the interface for layer options. | 17:43 |
rick_h_ | cory_fu: k, sounds good | 17:43 |
rick_h_ | cory_fu: did you see there's a charms and charming in the new discourse as well? | 17:43 |
rick_h_ | cory_fu: it'd be great to kick some discussions/etc that way as we flesh it out | 17:43 |
rick_h_ | bdx: have a test bundle I can demo today? | 17:44 |
cory_fu | rick_h_: Yep, I was working on a write up for the layer options bit on there already | 17:44 |
rick_h_ | cory_fu: <3 | 17:44 |
rick_h_ | cory_fu: kwmonroe bdx and anyone else that wants in, 10min to Juju Show | 17:50 |
rick_h_ | https://hangouts.google.com/hangouts/_/kstii25wdnd5jorqvle3grpblae for joining the conversation and ... | 17:51 |
rick_h_ | https://www.youtube.com/watch?v=R0R5DC7_Dio for watching me make a fool of myself live on the interwebs :) | 17:51 |
cory_fu | https://tutorials.ubuntu.com/tutorial/tutorial-charm-development-part1#0 | 18:02 |
kwmonroe | https://discourse.jujucharms.com/ | 18:03 |
cory_fu | I didn't realize you could bootstrap an older controller. That will be very useful | 18:10 |
cory_fu | https://docs.conjure-up.io/devel/en/conjurefile | 18:13 |
cory_fu | https://jujucharms.com/u/containers/aws-integrator/ | 18:14 |
cory_fu | https://jujucharms.com/u/containers/gcp-integrator/ | 18:14 |
cory_fu | https://jujucharms.com/u/containers/openstack-integrator/ | 18:14 |
cory_fu | Full changelog for conjure-up: https://github.com/conjure-up/conjure-up/blob/master/CHANGELOG.md | 18:17 |
rick_h_ | cory_fu: kwmonroe feel free to update https://discourse.jujucharms.com/t/juju-show-37-2-4-0-lxd-show-and-tell-and-more/60 if you have additional links/notes to drop in there | 18:29 |
rick_h_ | ty for the idea kwmonroe | 18:29 |
thumper | morning team | 20:32 |
dparrish | g'morning maestro! ;-) | 20:32 |
veebers | Morning all o/ | 20:53 |
magicaltrout | hacked together an oozie charm kwmonroe as i don't see one in the store anywhere thats current | 21:15 |
magicaltrout | i'll get someone to clean it up and send it upstream if you want to add it to the bigtop charms | 21:15 |
kwmonroe | +100 magicaltrout! i saw the open jira about puppet and oozie client/server. did you sort that out? | 21:17 |
magicaltrout | well | 21:18 |
magicaltrout | it's a bodge | 21:18 |
magicaltrout | but it works | 21:18 |
kwmonroe | i'll google bodge later | 21:18 |
magicaltrout | run puppet -> fix the package -> rerun puppet to finish the install | 21:19 |
kwmonroe | ah, so normal puppet then ;) | 21:19 |
magicaltrout | pretty much | 21:19 |
maaudet | What is supposed to show under the "LABELS" column when running juju metrics ? | 21:29 |
rick_h_ | maaudet: it's optional for charms to supply a label to the metric in order to tell things apart in the data coming out | 21:45 |
maaudet | rick_h_: I see. If I'd like to add a label to my metrics on my charms, where would I put it? Is it simply a key "label" to add in my metrics.yaml objects? | 21:49 |
rathore_ | rick_h_: Nice show today. For life of me I cannot get juju to work with bionic. The lxd containers have some issue with networking. The packet speeds are in bytes/sec. The configuration works well with xenial. https://paste.ubuntu.com/p/6YqtgNBpxZ/ | 21:49 |
veebers | wallyworld: I got access to the staging store, I needed to log in via the web ui, I needed to build charm from source and I also needed to RTFM for the image attach :-) | 21:53 |
wallyworld | veebers: great, so unblocked..... | 22:34 |
veebers | wallyworld: aye, onwards and upwards | 22:34 |
wallyworld | babbageclunk: i need a more experienced person to review a change with an upgrade step https://github.com/juju/juju/pull/8900 | 22:35 |
wallyworld | no rush | 22:35 |
veebers | wallyworld: just sorting out the test failures on my WIP PR, I think it'll need some more. If you get that chance could you review and assess?(https://github.com/juju/juju/pull/8896) | 22:35 |
wallyworld | sure | 22:35 |
wallyworld | veebers: quick initial comment - the controller config has the default value as "". It should probably be csclient.ServerURL. That avoids checks for != "" in places and also makes it explicit when printing controller config what the url used will be | 22:43 |
veebers | wallyworld: ah, good point. I though leaving it to the construction of the client and passing "" would default fine, but what you're suggesting is the same result && more explicit /obvious | 22:44 |
wallyworld | veebers: there's a place in the PR outside of the actual client where we check for != "" | 22:45 |
wallyworld | this would avoid that as well as being explicit | 22:45 |
veebers | ack | 22:53 |
wallyworld | veebers: left some initial comments | 23:02 |
veebers | wallyworld: ack. Oh, you dont' think conConfig doesn't just roll off the tongue? Like Con-Air? :-) | 23:05 |
wallyworld | vino: i left some comments - ping me if it's unclear | 23:50 |
vino | sure wallyworld | 23:51 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!