/srv/irclogs.ubuntu.com/2019/10/29/#juju.txt

kelvinliuwallyworld: got this pr to fix cm label config, +1 plz, thanks! https://github.com/juju/juju/pull/1081500:19
kelvinliulabel conflict00:21
wallyworldkelvinliu: can i get a review on this small fix? https://github.com/juju/juju/pull/1081401:22
kelvinliuwallyworld: yep, looking01:22
kelvinliuwallyworld: lgtm thanks 👍01:28
wallyworldthumper: by "k8s" workload, do you mean charmed kubernetes bundle deployed to the specified cloud?01:49
thumperyes01:49
thumperit is more than trivial but not extreme like openstack01:49
wallyworldi might clarify01:49
thumperok, thanks01:49
wallyworldalso, i think kubernetes-core bundle will do01:50
* thumper nods01:50
thumperthat deploys on lxd?01:50
wallyworldcharmed kubernetes takes twice as long to settle01:50
wallyworldjuist has HA componenets etc01:51
wallyworldyeah01:51
wallyworldlast time i looked which was ages ago01:51
wallyworldkubernetes-core is just a slimmed down charmed k8s01:51
babbageclunkthumper: got a bit fiddlier than expected - not sure what to put in QA steps without suggesting commenting out Unwatch calls from a watcher. https://github.com/juju/juju/pull/1081603:02
thumperbabbageclunk: thanks03:02
* thumper looks03:02
* thumper waves at alexisb03:43
hpidcockwallyworld: I just realised that changing how we handle actions on k8s will prevent the --operator functionality of juju exec while the workload is initialising, for example if it got stuck or took a long time this could be an issue. Also might make implementing hooks that run on the workload more difficult.04:12
hpidcockso I'm not sure the new solution is better than the one we already ahd04:12
wallyworldhpidcock: did you want a HO to discuss04:15
hpidcocksure04:15
anastasiamacwallyworld: thumper https://github.com/juju/juju/pull/1081705:06
wallyworldkelvinliu: this ensure all k8s resources have the expected juju annotations https://github.com/juju/juju/pull/1081805:30
kelvinliulooking05:31
wallyworldanastasiamac: i got a question, quick HO05:35
wallyworld?05:35
anastasiamack05:35
anastasiamacm back in 1:105:36
wallyworldanastasiamac: sorry, got to restart chrome, one sec05:36
anastasiamack05:36
kelvinliuwallyworld: so just want set annotations on all resources, but don't use them yet, right?05:41
wallyworldkelvinliu: nothing in code uses them. it's the same as for iaas. it's a way to mark resources created in the cloud as belonging to a particular controller, or adding user specified tags so they can look at a cloud dashboard and see certain annotations05:43
wallyworldor they might write a script to loist all resources with certain tags that match model config05:43
kelvinliuok, lgtm, thanks05:43
wallyworldtyvm05:43
wallyworldanastasiamac: +1 with fixes06:27
anastasiamacwallyworld: m about to commit another one.. and will address ur comments.. ta06:28
wallyworldanastasiamac: let me know if i'm on crack etc with anything06:29
wallyworldhopefully it all makes sense06:29
anastasiamacu mean with respect to comments or in general? :-P06:31
wallyworldboth06:32
wallyworldkelvinliu: any blockers with the custom resources? don't forget to include annotations :-) once my PR lands, you can copy of that06:36
kelvinliuwallyworld: still a bit early to see if there are any blockers..06:36
wallyworldno worries06:37
kelvinliubut got some ideas06:37
anastasiamacwallyworld: k, committed my last wish... looking at ur comment to determine the presence of crack in ur bllodstream06:50
wallyworldwe'll see...06:51
hpidcockwallyworld: still need to finish off unit tests, but you can take a look at the updated pr if you want https://github.com/juju/juju/pull/1080707:19
wallyworldok07:20
madsagegood morning07:24
anastasiamacwallyworld: addressed all... heading for landing and m loking at ur pr before/during/after dinner07:46
wallyworldty, can wait till tomorrow07:46
anastasiamac\o/ excellent I'll look tomorrow then :D07:46
wallyworldso long as we land before EOD07:47
anastasiamacyep... m hoping i'll land mine tonight so that u can rebase urs in the mornig and can review the rebased version :D07:47
wallyworldsgtm07:47
wallyworldi'll watch yours and rebase07:47
anastasiamac\o/07:47
nammn_demorning manadart: is the activeBranch a local concept? Or do we also save it on the jujud/mongo side?09:55
manadartnammn_de: Morning. It is local. See: ~/.local/share/juju/models.yaml09:56
nammn_demanadart: 🦸09:56
nammn_demanadart: juju/core/cache is jujud side cache, right?10:10
manadartnammn_de: It runs on controllers. The cached model is accessible from API server facades.10:11
nammn_demanadart: /cmd/* is everything related to the cli (juju) /apiserver/* is everything related to jujud agent (facade server, cache, ..)?10:16
manadartnammn_de: cmd has sub-packages for the CLI (juju) and agents (jujud). apiserver has sub-packages for facades that are accessible by clients/controllers/agents. The API server itself serves the facades, so the cached model is available to all of them.10:20
nammn_demanadart: uff, i need to let that sink and play a little. Thanks that helps a lot!10:21
nammn_dehey hml I tried to incorporate your suggestions. Was not 100% sure whats the best way as activebranch is a local information. https://github.com/juju/juju/pull/10796#discussion_r33923614712:20
hmlnammn_de:  rgr, i’ll take a look now12:20
hmlnammn_de: i was thinking that it wasn’t local only.  but it appears to be after all.  therefore the suggestion isn’t usable.  :-D12:25
=== fenris is now known as Guest55271
nammn_dehml: ok :D, but I still tried to incorporate it a little and added it to the branchstatus struct in formatter and thus could remove it from the other struct, so less usage12:56
hmlnammn_de: i saw.  working on qa now to answer some questions of my own.12:57
nammn_dehml: ahh sure, no pressure12:57
hmlnammn_de: i’m having not enought coffee syndrome trying to qa this morning.  sorry!12:58
nammn_derick_h: if you have some boring spare time : https://github.com/juju/juju/pull/10796 :D15:09
manadartnammn_de: Can I get you to QA this one? https://github.com/juju/juju/pull/10821/15:09
nammn_demanadart: can do15:09
hmlmanadart: what does colures mean how you’ve used in a commit msg? google gives me a defintion related to equinoxes/solstices.  :-)15:51
manadarthml: Which message? Might be a typo :)15:52
hmlmanadart: “Explicitly ignores errors for defered iterator colures in migration_import”    sha 92308d315:52
manadarthml: Meant "closures".15:53
hmlmanadart:  :-)15:58
hmlmanadart: reviewed 10821 left 2 comments.  approval tic is up to whomever QA’s.15:59
nammn_demanadart: how can I check the `machine address space IDs` ?16:00
manadartnammn_de: Connect to Mongo and run db.machines.find().pretty(). See https://discourse.jujucharms.com/t/login-into-mongodb/30916:01
manadartnammn_de: Appologies; too vague on QA steps.16:02
nammn_demanadart: does that mean, I bootstrap your code as maas1 -> save machineaddressspace; bootstrap your code as maas2; migrate mode maas1 to maas2 and check machineaddressspace of maas1 and maas2 are the same?16:05
nammn_demanadart: just read my test, its confusing. Here is mine in better: https://pastebin.canonical.com/p/kYxSKsKKx8/16:08
manadartnammn_de: This is from my history: https://pastebin.canonical.com/p/wrpRTSvxkS/16:10
nammn_demanadart: you my hero16:10
nammn_dethanks gonna test16:10
nammn_deim +1 for pasting history into qa steps, makes it for beginner, like me, easier16:11
nammn_de:D16:11
hmlrick_h: manadart  what do you think about renaming the model config “default-space” to “default-endpoint”?  or at least within the code.  It gets very confusing what the value is once you’re out of config code.16:29
rick_hhml:  the endpoint is the thing in the charm (http or such) so not sure about default-endpoint. If internally to the code something like default-space-for-binding or something makes more sense/use I could see it16:30
rick_hhml:  but it's a bit much for the operator in model-config imo, they get to cheat though and have a description and such in there16:30
hmlrick_h: right… just inside the code… will cause problems long term i think.  and get confused with other concepts16:31
rick_hhml: understand16:32
manadartrick_h hml: Now that there is no default for the model config value, I expect references to the default space name to dry up in favour of the config...16:46
rick_hmanadart:  right, but at least then there's one place to look and it'll notice when there is a value in place16:48
rick_hmanadart:  or am I missing a link?16:48
rick_hmanadart:  in the end the config is the "single source of truth" vs having a flag on the space struct "what's the default one from the list"? or really back to implementation detail16:49
hmlmanadart: it’s still confusing to have this thing config.DefaultSpace().  running around the code for now.16:49
hmlnot to be confused with what we’ve been calling the default space.  or empty space.16:50
hmlit can always be changed back once we’re gotten to the end of this project.16:51
manadartrick_h, hml: I can't see the issue. We now have the alpha/starter space, and the model config for the default.16:52
manadartJust change the current default consts. AlphaSpaceID, AlphaSpaceName or some such.16:53
nammn_demanadart: i did the qa, same version worked. Upgrade as well, but not sure about one output here: https://pastebin.canonical.com/p/PjNWFB7Vjd/ there is one space_id=116:53
hmlmanadart:  pondering16:54
manadartnammn_de: Yeah, if the instance poller runs, it will update the address space from MAAS. Probably better to have done it on another substrate.16:54
hmlmanadart:  i’ll try the AlphaSpaceName change, that might get us away from thinking that the space with ID 0 is a true default.16:58
bdxhello, I am working at the dc with our a few of our engineers today. We are trying to figure out how to enable juju to deploy workloads on a bare metal k8s, and having issues getting operator and workload storage configured correctly it seems. Wondering if anyone has experience in configuring workload and/or operator storage for a bare metal k8s?18:40
rick_hbdx:  I saw that folks were saying that they tend to do that with a file service on discourse but not setup on bare metal since the workloads are tied to the storage locations on the node18:59
bdxrick_h: I see19:01
bdxwe have the rbd provisioners configured https://paste.ubuntu.com/p/HsNjtNYrCy/19:03
bdxwe are trying to make sense of how to get juju use either of those storage pools for workload and operator storage19:04
bdxwe have followed this document through multiple times https://discourse.jujucharms.com/t/setting-up-static-kubernetes-storage-tutorial/119319:05
bdxas well as a few others19:05
bdxwe will be cycling on this today and the rest of the week, I'll report back with findings19:07
bdxthere seems to be a few good docs on juju/k8s storage out there, its just that none of them show the full picture (start to finish) of adding storage and getting a workload to run using that storage19:13
bdxI feel like we are really close ...19:13
bdxin getting hostpath and rbd provisioners to work19:13
bdxgoing to keep at it, can't be too far off19:14
=== Cynerva_ is now known as Cynerva
bdxheres what we did to get it working https://discourse.jujucharms.com/t/k8s-on-metal-storage-configuration-for-use-with-juju/2281/4?u=jamesbeedy21:50
anastasiamactimClicks: do we have 2.7-rc release notes or do I need to update existing 2.7-b123:01
anastasiamactimClicks: nm23:10
wallyworldhpidcock: can i get a review on this PR. it's a simple approach without requiring environ version (which gets messy and is a lot more work; i'd rather keep it up our sleeves for other things) https://github.com/juju/juju/pull/1081923:13
hpidcockNp, give me a seond, making coffee23:15
anastasiamacwallyworld: a trivial change, PTAL https://github.com/juju/juju/pull/1082623:33

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!