[01:18] thumper: https://github.com/juju/juju/pull/11609 [01:31] wallyworld: doing the rc1 upgrade I am seeing this in the controller after upgrade [01:31] ERROR juju.worker.dependency engine.go:671 "api-caller" manifold worker returned unexpected error: [bf8dc1] "controller-0" cannot open api: unable to connect to API: dial tcp 127.0.0.1:17070: connect: connection refused [01:31] any thoughts ? [01:33] wallyworld: tlm ho? [01:34] sure [01:34] k [01:58] wallyworld: lint failure [02:08] thumper: should be fixed [02:17] wallyworld: going trhough it now [02:28] wallyworld: bug 1878329 [02:28] Bug #1878329: stuck k8s workload unit following upgrade-charm with new image [02:29] tlm: did you want to chat? [02:30] thumper: in standup with wallyworld and kelvinliu discussing bugs atm [02:30] tlm: ok [02:30] tlm: I may leave you to work with them on the rc2 issues [02:30] no worries. Enjoy your arvo [02:30] I may give a brief try to see if I can reproduces the caas operater intermittent test failures locally [02:30] I have about 30 minutes [02:31] k, I am banging my head against the wall at the moment as I can do an rc1-rc2 upgrade. Juju doesn't bump the k8's image locally [02:31] seems to work for everyone else [02:39] tlm: running caas operator tests in a loop could reproduce in 400 iterations [02:39] but ran with -race and hit it in 10 [02:40] tlm: this line appears in failing test, but not passing test: [LOG] 0:00.087 WARNING test unit running chan["gitlab/0"] is blocked [02:43] hmmmm have you found where that log message comes from thumper ? [02:44] line above says: [02:44] // This should never happen unless there is a bug in the uniter. [02:44] caasoperator.go line 527 [02:44] can't get it to fail again [02:48] got it to fail eventually, but running stress-race and make check in another window [02:48] stress-ng doesn't seem to tickle it [02:48] I think I'd need to understand all the channels to debug further [02:48] * thumper reads code for 10 minutes [02:49] two ticks thumper, just writing a response in launchpad [02:52] free now thumper [02:54] tlm: which hangout? [02:54] standup ? [02:55] ack [03:42] kelvinliu: did you see paul has pasted links to the charm on the bug? [03:43] wallyworld: yep [03:43] i'm just landing some other stuff, will test soon [04:52] wallyworld, kelvinliu [04:52] whoops [04:53] PR https://github.com/juju/juju/pull/11610 no rush it's more of a look and see if you agree type situation [05:06] tlm: i'm always a little wary about creating a buffered channel to solve a race; in this case, I can't see from the code ho we can get events from the containerStartChan before we start selecting from the channel, so as yet i can't see how this solves the race... is there a causal link between the unbuffered chnnel and the race? [05:06] wallyworld: https://bugs.launchpad.net/juju/+bug/1879598 HO to discuss this issue? [05:06] Bug #1879598: terminating pod confuses k8s charm operator [05:07] sure one sec [05:09] kelvinliu: free now [05:10] stdup? [05:12] yup [05:36] wallyworld: I can try and explain it better over HO if you want [08:52] manadart, CR https://github.com/juju/juju/pull/11608 [08:55] stickupkid_: Swap you. https://github.com/juju/juju/pull/11607 [09:31] stickupkid_: Got a sec to HO? [09:31] sure [12:35] cory_fu: Know what is happening here? I think layer:status is broken with focal (py3.8), but I haven't traced why: https://pastebin.ubuntu.com/p/6bXFzkxB6b/ [12:36] manadart, when you get a second, want to talk through the next stages of work on bindings thing [12:37] cory_fu: Weirdly, hooks work fine, but actions failing (which use the charm-env shebang) [12:43] stickupkid_: Want to HO now before standup? [12:44] sure [12:48] cory_fu: nm, found charms.layer.basic.activate_venv where the docstring explains it [12:53] stub: I'm not sure about activate_venv, but I think actions need to explicitly call import_layer_libs: https://github.com/juju-solutions/layer-basic/blob/master/lib/charms/layer/__init__.py#L6 [12:54] cory_fu: activate_venv does that, and some other twiddling. 'This is handled automatically for normal hooks, but actions might need to invoke this manually,' [12:55] Ah, right, I see [12:56] charm-env does part of it by ensuring the venv is active, but not the extra twiddling [12:57] activate_venv works in any case \o/ . I have no idea when this broke; I haven't rebuilt the charm for ages. [12:57] Was thinking it was py3.8 problems [12:59] I think it's mostly due to the `from charms import layer` pattern vs `from charms.layer import status` === techalchemy_ is now known as techalchemy [15:18] manadart, if you're still around https://github.com/juju/juju/pull/11611 [15:19] manadart, this lifts up the endpoints binding calls along with AllSpaces call... [15:19] manadart, I'm going to tackle the merging in another PR [16:25] stickupkid_: I will have a proper look in the AM. Have to head off now. [16:25] manadart, wicked, nearly finished myself [22:25] What api facade/method can I use to get a controllers cloud/region ? [22:36] hatch: not sure about that, but show-controller has the details you need [22:36] check what facade it hits [22:37] thanks, I'll check there [22:44] all these side effects..... [22:57] thumper I'm not really sure where `details` is being populated https://github.com/juju/juju/blob/d59b8637e150943cb987a8194addbdc116746b63/cmd/juju/controller/showcontroller.go#L139