thumper | https://github.com/juju/juju/pull/10834/files | 00:49 |
---|---|---|
wallyworld | kelvinliu: +1 but with a request for validation, might be possible, let me know | 01:31 |
wallyworld | thumper: lgtm, ty | 01:32 |
kelvinliu | wallyworld: tyrv, no, we can't validate the cr before getting the CRD. because it's just a string for us without knowledge of the CRD. | 01:36 |
wallyworld | kelvinliu: yes, but can't we make an api call to validate? | 01:37 |
kelvinliu | but k8s will return the validation error if the deployed CR did't match the CRD format | 01:37 |
wallyworld | and we can't check that before actually trying to create one? | 01:38 |
wallyworld | there's no k8s Validate() method? | 01:38 |
wallyworld | i guess we can't be sure the crd exsists | 01:38 |
wallyworld | until be process the entire yaml | 01:38 |
kelvinliu | so just deploy, either success or fail(return validation error from k8s) would be better, | 01:38 |
wallyworld | what does kubectl do if the unstructured cr yaml is bad? | 01:39 |
kelvinliu | errors returns from api-server | 01:39 |
wallyworld | does it return an error or just do it and you see the error with kubectl get | 01:39 |
kelvinliu | we are following the kubectl way | 01:40 |
wallyworld | ok, so long as we set an error in juju status which i think we do | 01:40 |
kelvinliu | yes, we do | 01:40 |
wallyworld | ok, sgtm, ty | 01:40 |
kelvinliu | np | 01:44 |
wallyworld | hpidcock: getting through your PR, have a meeting in 5 but will continue after that | 01:52 |
timClicks | "Highlighting some important Juju workloads" https://discourse.jujucharms.com/t/2077 | 02:51 |
wallyworld | hpidcock: why do we need to annotate the pods with init id as part of the running check? | 02:52 |
wallyworld | ah, if you scale directly in k8s, the juju unit won't exist straight away | 02:56 |
timClicks | babbageclunk: nice work on the vsphere nics bug | 02:58 |
babbageclunk | :( | 02:59 |
babbageclunk | oops, thanks! | 02:59 |
hpidcock | wallyworld: yeah there is a race condition that the unit and provider id has not matched yet, which means the WatchContainerStart facade will filter it out because it doesn't know it yet. | 03:09 |
hpidcock | Also, it's now a nice useful debug annotation :) | 03:10 |
wallyworld | indeed | 03:11 |
hpidcock | anyone else getting test fails on localServerSuite.TestNetworkInterfacesForMultipleInstances? I think it's an ordering issue, but wondering if anyone has fixed it or seen it | 03:22 |
wallyworld | hpidcock: i've not seen it it myself. is it blocing landing your PR? | 04:56 |
hpidcock | I don't think it will block, something is using a map somewhere, it has 2 elements, so that makes it a 50% chance to land :D | 04:58 |
wallyworld | could be worse :-) | 04:58 |
wallyworld | kelvinliu: your PR good to land today too? | 04:59 |
kelvinliu | wallyworld: i think so, I m solving some races in tests. | 05:01 |
wallyworld | ok, ty! | 05:01 |
manadart | Still after a tick on https://github.com/juju/juju/pull/10821. | 08:20 |
manadart | nammn_de: Can you try the 2nd QA step on a different substrate for patch 10821 and approve if it passes? Need to get it landed. | 08:58 |
nammn_de | manadart: will do :-) | 09:01 |
nammn_de | does it have to be space aware like aws? Or does lxd work? | 09:02 |
manadart | nammn_de: LXD should work. | 09:05 |
achilleasa | manadart or nammn_de: one line PR for fixing the ec2 intermittent test failure that hpidcock reported: https://github.com/juju/juju/pull/10835 | 09:12 |
jam | manadart: have you considered whether we want to be passing space ids in the migration content? | 09:15 |
jam | manadart: that's the only thing keeping me from flagging that Approve, so if you're confident we can move forward | 09:15 |
manadart | jam: I am confident. I understand communicating in terms of space names from a conceptual point, but I don't think reverting from IDs gets us any more "correctness" for the effort. | 09:20 |
nammn_de | manadart: run qa, approved as you are confident and qa worked | 09:38 |
manadart | nammn_de: Ta. | 09:38 |
manadart | achilleasa: Looks like I need to rebase with your test fix for mine to play ball and land. | 10:37 |
achilleasa | manadart: fix is still landing... that test fails about 20% of the time so if you retry the merge it should probably work | 10:38 |
manadart | achilleasa: Giving it another lash. Last 2 failed. | 10:39 |
nammn_de | stickupkid: coming back to the charming thing. I was thinking a little how to make it consistent with the old behaviour while removing the bug we were talking before. I think this should do it https://github.com/juju/charm/pull/297/files | 10:44 |
stickupkid | nammn_de, nearly there :D | 11:05 |
achilleasa | manadart: fix has landed | 11:18 |
manadart | achilleasa: Yeah, mine was backed up behind yours, so I could have just waited. Running now. | 11:19 |
nammn_de | stickupkid: updated it | 11:50 |
stickupkid | nammn_de, much better | 11:51 |
stickupkid | readability ftw! | 11:52 |
nammn_de | stickupkid: let's hope big that now all things work as expected there compared with before :D | 11:53 |
manadart | achilleasa: Mine is in. Let me see if I can chase this issue. I think your time is better spent on the instance-poller stuff. | 12:01 |
hml | manadart: what do you mean by “a zero-value SpaceInfo “ for the peergrouper getHASpaceFromConfig() | 12:13 |
manadart | hml: "network.SpaceInfo{}" | 12:15 |
manadart | hml: Which is effectively sending "" for space ID/name. | 12:15 |
hml | manadart: ack | 12:16 |
achilleasa | hml: one small comment for your PR; I will run the QA steps in a bit | 12:47 |
hml | achilleasa: thank you. | 12:47 |
hml | achilleasa: if you have time after stand up, or after pick up, i’d like to chat on the other one pls 10825 | 12:48 |
achilleasa | hml: sure, I have 1h before pickup so let's do this after standup | 12:49 |
Mudchains | Hi all, I am deploy a kubernetes-core on a onpremise vpshere environment atm. its already 10 minutes 'stuck' at Juju Controller is initializing. Please wait. The console of the juju-controller VM is waiting at a logon screen. is this normal behavior? | 13:26 |
hml | manadart: ping? can you join daily for pr chat? | 13:40 |
manadart | hml: OMW | 13:43 |
nammn_de | manadart: do these changes make sense to you? I reuse the metrics pointer from application to use them in the charmconfigwatcher | 13:56 |
nammn_de | https://github.com/juju/juju/pull/10833 | 13:56 |
nammn_de | jam: I was looking into how some charm tests are written in juju/juju. Some are under /juju/juju/repositoy... and they contain a "version" file as well. Right now the "git describe dirty" would take precedence over the version file. Is that wanted? | 14:19 |
nammn_de | stickupkid: got a minute? | 14:41 |
nammn_de | stickupkid: https://github.com/juju/juju/pull/10836 added some background | 14:57 |
nammn_de | rick_h: above I sent to jam ^ should vcs or version file take priority? | 15:07 |
manadart | hml achilleasa: Should I expect all bindings not-explicitly defined to be "0" now? | 15:25 |
hml | manadart: yes, the exception is if the application’s default endpoint has a defined value. | 15:27 |
manadart | hml achilleasa: This looks like the issue with https://bugs.launchpad.net/juju/+bug/1850589 | 15:28 |
mup | Bug #1850589: migrating kubernetes-core fails <model-migration> <juju:Triaged by manadart> <https://launchpad.net/bugs/1850589> | 15:28 |
manadart | The bindings migrated from 2.6 are all "" in the 2.7 model. | 15:29 |
hml | manadart: i thought we had code to convert them when the bindings were migrated. hrm… | 15:31 |
achilleasa | manadart: hml That looks correct to me; it's the old default space name, no? | 15:40 |
hml | achilleasa: yes… but we should be saving in to the db with an ID, not a name | 15:41 |
hml | achilleasa: manadart i think i found a bug in the cmr code related to the space name/id… spaceIDs do not translate outside of the given model. spaces are used to match up subnets on each side of the offer. | 15:45 |
Mudchains | Ok it was defently a firewall issue. conjure-up is now deploying the kubernetes-core on the vsphere environment | 15:45 |
manadart | achilleasa hml: Look at what I posted in the bug. The issue is here: https://github.com/juju/juju/blob/6bf5de9ded487a4ecb8d23a8118ef3df030c24c0/state/migration_import.go#L853 | 15:48 |
manadart | We are relying on NewBindings to return endpoint:space-id, but if all are "", it returns the original map. | 15:49 |
manadart | So we import it untouched from an old controller with no explicitly bound endpoints. | 15:50 |
achilleasa | hml: this is why we should probably dumb down NewBindings so it only works with space ids... ^^ | 16:00 |
hml | achilleasa: ensuring it’s called that way might be a bigger problem. and changing code handingly “” | 16:02 |
manadart | achilleasa hml: Who should I assign it to? I have to make a move here. | 16:06 |
hml | achilleasa: can you pick it up? i’m knee deep in get the other two PRs going | 16:06 |
achilleasa | hml: on it | 16:24 |
hml | achilleasa: awesome! | 16:24 |
achilleasa | hml: ok, so the issue here is that when you export from 2.6.x the description contains ep => space name (with "" for the default) whereas for 2.7+ it will contain space IDs... I guess the fix is to auto-map "" => AlphaSpaceID | 16:38 |
hml | achilleasa: yes, a good place to start. we should perhaps review with handing of model config default-space. | 16:39 |
achilleasa | hml: since the model config is a 2.7 only feature, I think that the migration should not take it into account as it could lead to unpredictable bindings | 16:52 |
hml | achilleasa: gd point | 16:53 |
nammn_de | Someone wanna take a look at this one? https://github.com/juju/juju/pull/10838 Adds bash branch completion | 17:20 |
rick_h | nammn_de: sure thing | 17:22 |
achilleasa | rick_h: should 'juju run --unit ubuntu-lite/0 "network-get another"' print something out? | 17:59 |
rick_h | achilleasa: no? since it's not in a relation context? | 18:01 |
rick_h | achilleasa: or maybe...thinking more | 18:01 |
achilleasa | rick_h: ah yes... | 18:01 |
* rick_h checks if that's an extra-binding or a endpoint | 18:01 | |
achilleasa | well, I have a fix for thumper's issue and probably https://bugs.launchpad.net/juju/+bug/1850589 | 18:02 |
mup | Bug #1850589: migrating kubernetes-core fails <model-migration> <juju:In Progress by achilleasa> <https://launchpad.net/bugs/1850589> | 18:02 |
rick_h | achilleasa: hmm, what does 2.6 do with it? | 18:03 |
achilleasa | rick_h: something odd is going on 'github.com/juju/juju/cmd/juju/commands/exec.go:434: timed out waiting for result from: unit ubuntu-lite/0' | 18:26 |
rick_h | achilleasa: yea... | 18:27 |
rick_h | achilleasa: working to rebootstrap/test that here as well | 18:27 |
timClicks | have a minute? any eyes on this piece is appreciated: https://discourse.jujucharms.com/t/draft-juju-2-7-0-creating-a-new-usability-standard-for-infrastructure-automation/2294 | 21:55 |
=== narindergupta is now known as narinderguptamac | ||
anastasiamac | fwiw, we already collect and display in machine-format cloud 'source' | 23:37 |
anastasiamac | wallyworld: path of less resistence is to add it to tabular as 'Kind' | 23:38 |
anastasiamac | wallyworld: but 'personal' clouds r called 'local' atm.. m happy to keep them as 'local'.. r u? | 23:38 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!