[01:53] <axw> wallyworld: I put it on the calendar already, but just reminding you: I'm away this coming Fri-Mon
[02:03] <axw> babbageclunk: do you have time to be a second reviewer for https://github.com/juju/juju/pull/7984?
[02:04] <babbageclunk> axw: sure
[02:04] <babbageclunk> axw: net line reduction automatic approval
[02:05] <axw> :)
[02:05] <babbageclunk> I mean, ok, I'll actually read it
[03:33] <axw> wallyworld: what version of Juju was the relation bug found in? beta1? I'm failing to repro with wordpress-mysql. just noticed bug says mediawiki, will try that - can't see why it'd matter tho
[03:39] <thumper> https://github.com/juju/juju/pull/7987 fixes constraints handling with bundle deployment.
[03:39] <wallyworld> axw: i tested in tip of develop last week. found it by remove-relation <id> from the offering model, and then add-relation again from the consuming model
[03:40] <wallyworld> axw: i'll try and repro
[03:41] <axw> wallyworld: ah, I just removed it from the consuming model. I'll try that way
[03:41] <babbageclunk> wallyworld: one of the migration test failures looked like a timeout, the other one looks like a genuine problem - we require a model type when importing a model, but older versions don't set one when exporting.
[03:41] <wallyworld> ah
[03:41] <wallyworld> i guess default to IAAS
[03:41] <babbageclunk> creating a bug now
[03:41] <babbageclunk> yu[
[03:41] <babbageclunk> p
[03:41] <veebers> babbageclunk: I don't think it's a timeout as the whole test is about 15 minutes, timeout is 50m. That being said I'm unsure why that log just stops
[03:42] <wallyworld> veebers: you're back!
[03:42] <babbageclunk> veebers: hmm
[03:42] <veebers> wallyworld: hey o/ aye I am, first day back today :-)
[03:42] <axw> wallyworld: seems to be some other bug though. I removed relation between hosted mysql and wordpress. something caused wordpress unit to restart, and now it's in an error loop complaining about the relation directory being non-empty
[03:43] <wallyworld> interesting. i wonder how mediawiki behaves
[03:43] <veebers> babbageclunk: let me have another quick look at those logs
[03:45] <wallyworld> thumper: lgtm
[03:46] <axw> wallyworld: remove-relation on the offering side didn't trigger anything. I'll look into the other thing I found
[03:46] <wallyworld> righto, i'll try and reproduce
[03:46] <axw> I mean didn't trigger any errors... re-adding relation just worked
[03:46] <wallyworld> axw: you tested with lxd?
[03:47] <axw> wallyworld: yep
[03:52] <axw> wallyworld: and now the remote relation worker seems to have stopped working: ERROR juju.worker exited "mysql": watching status for offer: verification failed: macaroon not found in storage
[03:52] <axw> "juju offers" shows no relation, though the client thinks it has related
[03:53] <wallyworld> axw: hmmm, looks like macaroon has got cleaned up on offer side but we are not minting a new one
[04:00] <wallyworld> axw: i have a meeting now but will debug a bit when i am free
[04:00] <axw> wallyworld: sure. I'll keep digging
[04:01] <thumper> wallyworld: there in a minute
[04:02] <axw> so if I remove the remote app from the consuming side, and then re-relate, it fixes it up
[04:05] <wallyworld> axw:  that would be because that causes a RegisterRelation which mints a new macaroon
[04:08] <axw> wallyworld: the remote relation worker should be re-registering if the relation was removed. I suppose the worker isn't tidying up properly. will look more after lunch
[04:09] <wallyworld> axw: yeah, i'll look between meetings if i get time
[04:30] <babbageclunk> axw: fixing this - https://bugs.launchpad.net/juju/+bug/1728486
[04:30] <mup> Bug #1728486: migration: allow model type to be blank when importing <juju:Triaged by 2-xtian> <https://launchpad.net/bugs/1728486>
[04:30] <babbageclunk> axw: I should obviously make the reading side accept blank.
[04:31] <babbageclunk> axw: Do you think I should make sure the exporting side also sets model type?
[04:32] <babbageclunk> axw: I think the intention was that the 2.2 branch would export models as v3, so type would be set automatically.
[04:32] <axw> babbageclunk: sorry just looking at the code, one minute
[04:33] <babbageclunk> axw: but the description dependency was updated in 2.2 for some unrelated change and now it emits v4 models, but with blank type.
[04:33] <babbageclunk> axw: thanks
[04:35] <axw> babbageclunk: ugh, that's unfortunate. yes I think the exporter should be required to set the model type for v4
[04:35] <axw> babbageclunk: reading side shouldn't have needed to accept blank, but obviously will now
[04:36] <babbageclunk> axw: do you think I should set the default in description or in migration_export.go?
[04:38] <axw> babbageclunk: I think migration_export.go, I think description should be relatively dumb - not making choices, just checking the choices you made are correct
[04:39] <babbageclunk> axw: cool, thanks!
[05:13] <thumper> later peeps
[05:13]  * thumper EODs
[05:29] <wallyworld> jam: did you forget to add a video call?
[05:30] <jam> wallyworld: I did, but I just added "caas"
[05:30] <jam> I will brt, need to use a restroom
[07:36] <axw> wallyworld: the unitcount refcounting on relations for local vs. remote units looks dodgy. we never increment for remote units, but we sometimes decrement for remote
[07:36] <axw> that might explain the "permission denied" error I saw
[08:41] <axw> wallyworld: I think we should be maintaining relation.unitcount for both local and remote units. i.e. unitcount == number of units in scope, at all times
[08:41] <axw> wallyworld: otherwise you can remove the relation without taking units out of scope
[13:52] <axw> wallyworld: why does having a time-before caveat stop things from working? you should still get a verification error?
[15:23] <mup> Bug #1726945 changed: [2.3] consume feature hard to manage <docteam> <juju-core:Invalid> <https://launchpad.net/bugs/1726945>
[15:29] <mup> Bug #1726945 opened: [2.3] consume feature hard to manage <docteam> <juju-core:Invalid> <https://launchpad.net/bugs/1726945>
[15:39] <mup> Bug #1726945 changed: [2.3] consume feature hard to manage <docteam> <juju-core:Invalid> <https://launchpad.net/bugs/1726945>
[21:12] <bdx> heyyyyy
[21:12] <bdx> I've got this model in a crazy stuck state on 2.3-beta2.5
[21:12] <bdx> lol
[21:12] <bdx> 2.3-beta2.1
[21:13] <bdx> whoops
[21:13] <bdx> ha
[21:13] <bdx> not sure if anyone wants to take a gander
[21:13] <bdx> http://paste.ubuntu.com/25854148/
[21:14] <bdx> I can't get the units es-lxd-test/{0,1} to go away
[21:14] <bdx> I mean, its not really a big deal to me
[21:14] <bdx> but, possibly its a difficult situation to repro
[21:15] <bdx> and someone might want to hop in and take what they need
[21:15] <bdx> I'll leave it up for the next day or so
[21:15] <bdx> if anyone wants to take a closer look, ping me
[21:15] <bdx> I'll create a bug anyway
[21:17] <bdx> well, not that tough
[21:19] <bdx> https://bugs.launchpad.net/juju/+bug/1728721
[21:19] <mup> Bug #1728721: units stuck in model after failed deploy <juju:New> <https://launchpad.net/bugs/1728721>
[21:29] <hml> bdx: do you have a build number for your juju 2.3-beta2.1?
[21:33] <bdx> hml: its whats in the beta snap channel
[21:33] <bdx> rev: 2640
[21:33] <hml> bdx: ty